Information processing system

Abstract

This invention provides an information processing system. The information processing system is characterized by comprising: a processor; wherein the processor is configured to: receive information from a user and save the information to a database; parse the received information using a generative artificial intelligence model, extracting important keywords from the information; generate prompts based on the extracted keywords for selecting the most suitable financial assistance scheme; and automatically generate application documents for applying for the financial assistance scheme using the prompts generated based on the requirements of the selected financial assistance scheme.

Description

Technical Field

[0001] The technology disclosed herein relates to an information processing system. Background Technology

[0002] Japanese Patent Application Publication No. 2022-180282 discloses a method for controlling a role-based chatbot executed by at least one processor. The method includes the following steps: receiving a user's speech; adding the user's speech to a prompt word, the prompt word containing instruction statements associated with an explanation of the chatbot's role; encoding the prompt word; and inputting the encoded prompt word into a language model to generate a chatbot response to the user's speech.

[0003] In existing technologies, users applying for various financial aid or subsidy programs typically need to manually search for and understand complex program information, application conditions, and required materials. Users must manually analyze their own projects or needs and compare them with multiple financial aid programs to determine applicability. This process is not only time-consuming and labor-intensive but also prone to errors due to misunderstandings or omissions, leading to unsuitable programs, incomplete application materials, or inappropriate expression, thus reducing the success rate. Furthermore, even with the introduction of generative artificial intelligence technology, existing solutions often remain at the level of simple text generation, failing to form an integrated process from information reception and storage, keyword extraction, program matching, automatic application document generation, to user interaction and correction. In addition, existing systems often ignore the impact of user emotional states on the expression of application content and interactive experience, lacking a mechanism for flexibly adjusting application documents based on user emotions. Therefore, how to provide an integrated system that can utilize generative artificial intelligence to deeply analyze user information, automatically select the most suitable financial aid program, efficiently generate application documents, and support user emotion perception and interactive correction is the technical challenge that this invention aims to address. Summary of the Invention

[0004] To address the aforementioned issues, this invention proposes an information processing system comprising a processor. The processor is configured to receive information from a user and save it to a database, thereby enabling persistent storage and subsequent reuse of the user's original input. The processor utilizes a generative artificial intelligence model to parse the received information, extracting key keywords and generating prompts based on these keywords to select the most suitable financial assistance scheme. Using these prompts, the processor can intelligently match and filter within a pre-built set of financial assistance schemes, automatically determining the scheme most suitable for the user's needs. Furthermore, the processor uses the prompts generated based on the requirements of the selected financial assistance scheme to automatically generate application documents for applying for that scheme, achieving efficient and standardized generation of application documents. Preferably, the processor is also configured to analyze the user's emotional state and generate prompts based on this state to adjust the application documents, thereby adaptively adjusting the wording, tone, and length to improve user experience and document expression. Furthermore, the processor is configured to display the generated application documents through a user interface, allowing users to confirm and correct them. This creates a closed-loop process combining automatic generation and manual correction, ensuring that the final submitted application documents not only meet the formal requirements of the funding assistance system but also accurately reflect the user's true intentions and project content. Through the above structural and functional configuration, this invention effectively simplifies the funding assistance application process, reduces the burden on users, and increases the application success rate.

[0005] A "system" refers to an overall device or platform consisting of one or more hardware components and software modules, used to perform a series of processing steps such as information reception, storage, parsing, matching of financial assistance systems, and generation and display of application documents. It can be implemented in the form of a server, cloud platform, local device, or a combination thereof.

[0006] A "processor" is a computing unit that can execute program instructions, process input data and output processing results. It can be a single physical processor, a combination of multiple processors, or a processing module composed of CPU, GPU, dedicated AI chip, etc. It also includes logical processing units implemented in the form of distributed computing.

[0007] "User" refers to the entity that uses the system to submit information and request recommendations for financial assistance and the generation of application documents. It can be a natural person, legal person, or other organization, such as an individual, enterprise, public institution, or local autonomous body.

[0008] "Information" refers to content related to a funding application that a user enters or provides through the system, including but not limited to project background, project purpose, implementation details, budget, timeline, geographical location, organizational structure, and other textual, numerical, or structured data.

[0009] A database is a logical or physical collection used to store and manage data. It can store user information, parsing results, keywords, financial assistance system information, and generated application documents, and can be implemented using relational databases, non-relational databases, or other data storage media.

[0010] "Generative AI models" refer to AI models that can generate text, prompts, or other content based on input data, including but not limited to large language models, text generation models, or deep learning models with natural language understanding and generation capabilities.

[0011] "Parsing" refers to the process by which a processor uses generative artificial intelligence models or other algorithms to analyze and understand information provided by a user, including operations such as word segmentation, semantic understanding, entity recognition, topic analysis, and structured extraction.

[0012] "Keywords" refer to important words or phrases extracted from user information that can represent the core content of the information or are related to the financial assistance system, and are used for subsequent selection of financial assistance system and construction of prompt words.

[0013] "Financial assistance system" refers to various systems or programs established by government agencies, public organizations, enterprises or other institutions to provide financial support for specific projects, activities or entities, including subsidies, grants, severance pay, and awards.

[0014] "Prompt words" refer to text input or instructions used to guide generative artificial intelligence models to perform specific task generation or reasoning. In this invention, they are used to select funding assistance systems and automatically generate or adjust application documents.

[0015] "Requirements" refer to the application conditions and formal requirements stipulated by each funding assistance system, including the applicable targets, support areas, implementation regions, budget scope, types of materials to be submitted, document formats, etc., which are used to determine whether the application meets the requirements of the system and generate the content of the application documents accordingly.

[0016] "Application documents" refers to a collection of documents or electronic files generated for applying for a specific funding assistance program, including application forms, plans, budget sheets, instructions, etc., and the content should include at least a project overview, information on the implementing entity, funding structure, and other information required by the program.

[0017] "Emotional state" refers to the user's emotional or psychological tendency during the use of the system, inferred from the user's input text, interaction behavior, or other available information. Examples include tension, anxiety, optimism, and neutrality. It is used to adjust the language style or interaction method of application documents.

[0018] "Adjustment" refers to the modification or optimization of the content, structure, or expression of application documents based on the user's emotional state or other conditions. This includes modifying wording, changing tone, adding or deleting explanatory content, etc., to improve the readability and suitability of the documents.

[0019] "User interface" refers to the interface form used for information display and interaction between humans and machines. It can be a web page interface, mobile application interface, desktop application interface or other graphical / textual interface, used to display application documents and receive user confirmation and correction input.

[0020] "Display" refers to the process of presenting the content of generated or stored application documents to users in a visual form through a user interface, including displaying interface elements such as text, tables, and buttons on the screen.

[0021] "Confirmation" refers to the user's action of checking the generated application documents and expressing approval through the user interface, which can be achieved by clicking the confirmation button, the submit button, or other interactive methods.

[0022] "Correction" refers to the user's operation of modifying, supplementing, deleting, or replacing the content of the generated application documents in the user interface, so that the final application documents more accurately reflect the user's true intentions and the actual situation of the project. Attached Figure Description

[0023] Figure 1 This is a conceptual diagram illustrating an example of the configuration of the data processing system according to the first embodiment.

[0024] Figure 2 This is a conceptual diagram illustrating an example of the main functions of the data processing apparatus and smart device according to the first embodiment.

[0025] Figure 3 This is a conceptual diagram illustrating an example of the configuration of the data processing system according to the second embodiment.

[0026] Figure 4 This is a conceptual diagram illustrating an example of the main functions of the data processing device and smart glasses according to the second embodiment.

[0027] Figure 5 This is a conceptual diagram illustrating an example of the configuration of the data processing system according to the third embodiment.

[0028] Figure 6This is a conceptual diagram illustrating an example of the main functions of the data processing apparatus and head-mounted terminal according to the third embodiment.

[0029] Figure 7 This is a conceptual diagram illustrating an example of the configuration of the data processing system according to the fourth embodiment.

[0030] Figure 8 This is a conceptual diagram illustrating an example of the main functions of the data processing device and robot according to the fourth embodiment.

[0031] Figure 9 This represents an emotion map that maps multiple emotions.

[0032] Figure 10 This represents an emotion map that maps multiple emotions.

[0033] Figure 11 This is a sequence diagram illustrating the processing flow of the data processing system of the first embodiment.

[0034] Figure 12 This is a sequence diagram illustrating the processing flow of the data processing system in Application Example 1.

[0035] Figure 13 This is a sequence diagram illustrating the processing flow of the data processing system of the second embodiment.

[0036] Figure 14 This is a sequence diagram illustrating the processing flow of the data processing system in Application Example 2. Detailed Implementation

[0037] Hereinafter, an example of an implementation of the system to which the technology of this disclosure relates will be described with reference to the accompanying drawings.

[0038] First, let me explain the terminology used in the following instructions.

[0039] In the following embodiments, the processor (hereinafter referred to as "processor") with reference numerals may be a single computing device or a combination of multiple computing devices. Furthermore, the processor may be a single computing device or a combination of multiple computing devices. Examples of computing devices include CPU (Central Processing Unit), GPU (Graphics Processing Unit), GPGPU (General-Purpose computing on Graphics Processing Units), APU (Accelerated Processing Unit), etc.

[0040] In the following embodiments, RAM (Random Access Memory), as indicated in the figures, is a memory that temporarily stores information and is used as working memory by the processor.

[0041] In the following embodiments, the memory, as indicated by the reference numerals, is one or more non-volatile storage devices that store various programs and parameters. Examples of non-volatile storage devices include flash memory (SSD (Solid State Drive)), disks (e.g., hard disks), or magnetic tapes.

[0042] In the following embodiments, the communication I / F (Interface) with reference numerals is an interface that includes a communication processor and an antenna, etc. The communication I / F is responsible for communication between multiple computers. As an example of a communication specification applicable to the communication I / F, wireless communication specifications such as 5G (5th Generation Mobile Communication System), Wi-Fi (wireless fidelity) (registered trademark), or Bluetooth (registered trademark) can be listed.

[0043] In the following embodiments, "A and / or B" is synonymous with "at least one of A and B". That is, "A and / or B" means that it can be only A, only B, or a combination of A and B. Furthermore, in this specification, when "and / or" connects to express more than three items, the same interpretation as "A and / or B" applies.

[0044] First Implementation Method Figure 1 An example of the configuration of the data processing system 10 according to the first embodiment is shown.

[0045] like Figure 1 As shown, the data processing system 10 includes a data processing device 12 and an intelligent device 14. A server can be cited as an example of the data processing device 12.

[0046] The data processing apparatus 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" as understood in this disclosure. The computer 22 includes a processor 28, RAM 30, and memory 32. The processor 28, RAM 30, and memory 32 are connected to a bus 34. Furthermore, the database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0047] The smart device 14 includes a computer 36, a receiving device 38, an output device 40, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, RAM 48, and memory 50. The processor 46, RAM 48, and memory 50 are connected to a bus 52. In addition, the receiving device 38, output device 40, camera 42, and communication I / F 44 are also connected to the bus 52.

[0048] The receiving device 38 includes a touchscreen 38A and a microphone 38B, and receives user input. The touchscreen 38A receives user input via touch by detecting contact with an indicator (e.g., a pen or finger). The microphone 38B receives user input via sound by detecting the user's voice. The control unit 46A in the processor 46 sends data representing the user input received by the touchscreen 38A and microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the data representing the user input.

[0049] The output device 40 includes a display 40A and a speaker 40B, and presents data to the user 20 by outputting data in a form perceptible to the user 20 (e.g., sound and / or text). The display 40A displays visual information such as text and images according to instructions from the processor 46. The speaker 40B outputs sound according to instructions from the processor 46. The camera 42 is a small digital camera equipped with an optical system such as a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor.

[0050] Communication I / F44 is connected to network 54. Communication I / F44 and 26 are responsible for sending and receiving various information between processor 46 and processor 28 via network 54.

[0051] Figure 2The diagram shows an example of the main functions of the data processing device 12 and the smart device 14.

[0052] like Figure 2 As shown, in the data processing apparatus 12, specific processing is performed by the processor 28. A specific processing program 56 is stored in the memory 32. The specific processing program 56 is an example of a "program" as understood in this disclosure. The processor 28 reads the specific processing program 56 from the memory 32 and executes the read specific processing program 56 on the RAM 30. Specific processing is implemented by the processor 28 operating as a specific processing unit 290 according to the specific processing program 56 executed on the RAM 30.

[0053] The memory 32 stores a data generation model 58 and an emotion-specific model 59. The data generation model 58 and the emotion-specific model 59 are used by the specific processing unit 290. The specific processing unit 290 can use the emotion-specific model 59 to infer the user's emotions and perform specific processing based on the user's emotions. In the emotion inference function (emotion-specific function) using the emotion-specific model 59, various inferences and predictions related to the user's emotions are performed, including inferences and predictions of the user's emotions, but this is not limited to this example. Furthermore, emotion inference and prediction may also include, for example, emotion analysis (parsing).

[0054] In the smart device 14, the processor 46 performs the acceptance output processing. The memory 50 stores the acceptance output program 60. The acceptance output program 60 is used in conjunction with the data processing system 10 and the specific processing program 56. The processor 46 reads the acceptance output program 60 from the memory 50 and executes the read acceptance output program 60 on the RAM 48. The acceptance output processing is implemented by the processor 46 acting as the control unit 46A according to the acceptance output program 60 executed on the RAM 48. Furthermore, the smart device 14 has the same data generation model and emotion-specific model as the data generation model 58 and the emotion-specific model 59, and these models can also be used to perform the same processing as the specific processing unit 290. The acceptance output processing is implemented by the processor 46 acting as the control unit 46A according to the acceptance output program 60 executed on the RAM 48.

[0055] Alternatively, other devices besides the data processing device 12 may also have the data generation model 58. For example, a server device (e.g., a generation server) may have the data generation model 58. In this case, the data processing device 12 obtains the processing results (prediction results, etc.) using the data generation model 58 by communicating with the server device that has the data generation model 58. Furthermore, the data processing device 12 may be a server device or a user-held terminal device (e.g., a mobile phone, robot, home appliance, etc.). Next, an example of the processing of the data processing system 10 of the first embodiment will be described.

[0056] Example 1 The flow of a specific process in Example 1 will be described. Each part of the system described below is implemented by the data processing device 12 and the smart device 14. Furthermore, the data processing device 12 is referred to as the "server," and the smart device 14 is referred to as the "terminal."

[0057] Existing technologies for assisting financial aid applications typically rely solely on rule matching or fixed keyword retrieval to help applicants select aid programs, generating application documents using static templates. This approach presents several problems at the computer technology level: First, the server-side has limited processing capabilities for natural language business content, only capable of simple string matching and field filling. It cannot utilize generative AI models to perform semantic analysis of long texts, automatically extract business classification information and constraints, resulting in low accuracy in program matching, requiring significant manual intervention, and inefficient use of computational resources. Second, the interaction between the server and the generative AI model is usually a one-time, fixed-prompt call, unable to dynamically generate and combine prompts based on the different needs of the parsing, evaluation, and document generation stages. This leads to scattered and unreusable model inference results, increasing application-layer logic complexity and latency. Third, when generating application documents, most systems treat templates as static files and perform simple placeholder replacements at the application layer. They cannot structurally link the results of generative AI models with subsequent user interactions and modifications, nor can they progressively regenerate or complete parts of the document. This results in a large amount of redundant server-client round-trip communication, impacting overall system throughput and response time. Fourth, existing systems only provide simple text input and preview functions at the user interface layer, lacking item-by-item editing control and version management mechanisms linked to the server-side document structure model. They cannot utilize generative AI models to perform fine-grained re-analysis and local updates of user input, making it difficult to efficiently support the editing of complex application documents. Fifth, in terms of user personalization, traditional systems typically only make simple prompts or format adjustments on the front end. The server does not take user interaction behavior and emotional state as input signals and cannot adaptively adjust prompts and document expression styles at the model invocation level. This prevents the system from optimizing user interaction paths and server resource scheduling while ensuring the correctness of structured data.

[0058] Therefore, a system architecture is needed that integrates deep semantic parsing of natural language business content, dynamic prompt generation, system matching and scoring, templated document generation, and user interaction-driven partial regeneration control on the server side. This architecture aims to improve the way the server calls generative artificial intelligence models, manages data structures, and interacts with the human-computer interface from a computer technology perspective, thereby enhancing the automation, accuracy, resource utilization efficiency, and overall system response performance of financial assistance procedures.

[0059] The specific processing performed by the specific processing unit 290 of the data processing apparatus 12 in Embodiment 1 is achieved by the following means.

[0060] In this invention, the server includes: a processing unit for acquiring business content information in natural language form from a user terminal and storing the business content information as structured application input information in a data storage device; a processing unit for dynamically generating parsing prompts for a generative artificial intelligence model using the business content information as input, and calling the generative artificial intelligence model to perform natural language processing on the business content information to obtain parsing results including business classification information, target domain information, target subject information, and region information; and a processing unit for performing a retrieval on an information storage device storing financial assistance system information based on the parsing results, extracting multiple candidate financial assistance systems while considering system applicability information and target domain information, and performing fit calculations on the multiple candidate financial assistance systems using evaluation prompts for a generative artificial intelligence model when necessary. The system includes: a processing unit for scoring based on a specific optimal funding assistance system; a processing unit for acquiring document style information corresponding to the specific funding assistance system, generating a document generation prompt statement for automatically allocating the parsing result and a portion of the business content information to placeholders in the document style information, and automatically generating structured application document data based on the document generation prompt statement; a processing unit for converting the application document data into display data and sending it to the user interface, receiving additional input and modification information from the user, and updating the application document data and the application input information in the data storage device accordingly; and a processing unit for sending the application document data as an electronic application to an external processing device or external acceptance device based on the updated application document data and the sending destination information associated with the specific funding assistance system, and recording the sending result as management information. This allows for a multi-stage prompt-driven computation process within the server, centered on a generative artificial intelligence model. This enables automatic semantic parsing of natural language business content, accurate matching and scoring of financial aid systems, automatic generation of templated application documents, and local regeneration based on user interaction. While reducing manual intervention and unnecessary network round trips, this improves the structuring and reusability of server-side data processing, enhances the computational efficiency and response performance of generative artificial intelligence model calls, and thus improves the overall technical performance of the computer system in the field of financial aid application processing.

[0061] "User terminal" refers to information processing equipment operated by users for inputting business content information, viewing and editing application documents, and sending and receiving data to and from the server, including but not limited to computer equipment, mobile terminal equipment, or other electronic devices with network communication and display functions.

[0062] "Business content information in natural language form" refers to a collection of business information related to funding assistance applications, such as project overview, project purpose, implementation content, budget, target audience, and implementation area, described by users in natural language text.

[0063] "Structured application input information" refers to application data represented by predefined fields or data structures, obtained by the server after formatting business content information in natural language form. This includes structured data that can be directly processed by computer programs, such as project name field, project category field, budget field, and region field.

[0064] "Data storage device" refers to data storage resources used to store structured application input information, financial assistance system information, application document data and related management information in a readable and writable manner, including database systems, file storage systems or other non-transitory computer-readable storage media.

[0065] "Generative AI model" refers to an AI model that can generate text output or structured data output based on input prompts. It is trained on large-scale data, has natural language understanding and generation capabilities, and can be used to parse business content information, calculate suitability scores, and generate application document content.

[0066] "Prompt statements" refer to the instructional text or data that the server constructs and inputs into the generative artificial intelligence model when calling it. These instructions are used to define the model's role, task, output format, and processing scope, in order to guide the model to parse, evaluate, or generate corresponding results from business content information or institutional information.

[0067] "Parsing prompt statements" refer to prompt statements constructed with the purpose of extracting business classification information, target domain information, target subject information, and regional information from business content information. They are used to guide generative artificial intelligence models to perform semantic parsing on natural language text and output structured parsing results.

[0068] "Business classification information" refers to the identification information determined based on semantic analysis of business content information, used to indicate the category or type to which a business belongs, including but not limited to classification labels indicating the nature of the business such as "tourism promotion", "local revitalization" and "R&D support".

[0069] "Target domain information" refers to the identifying information extracted from the analysis of business content information to represent the technical, industrial, or policy domains involved in the business, such as "tourism domain", "education domain", "environmental protection domain", etc.

[0070] "Target subject information" refers to the identification information extracted from business content information to represent the type of business implementation object or beneficiary, including abstract subject categories such as enterprises, groups, individuals, and local institutions.

[0071] "Regional information" refers to the identification information related to the location of business implementation or the applicable geographical scope, which is parsed from the business content information, such as administrative regions, regional levels, or geographical attribute information.

[0072] "Information on financial assistance systems" refers to a collection of data stored in an information storage device, including the system name, applicable conditions, target beneficiaries, applicable fields, subsidy amount range, application period, and required document types for various financial assistance systems.

[0073] "Information on the conditions for the application of a system" refers to the conditional data in the information on financial assistance systems used to determine whether a specific system is applicable to a certain business, including conditions on the type of recipient, industry sector, region, budget size, and time.

[0074] "Candidate financial aid schemes" refers to a set of multiple financial aid schemes that are potentially applicable and are selected by the server during the retrieval phase based on the comparison between the analysis results and the financial aid scheme information.

[0075] "Evaluation prompt statements" refer to prompt statements constructed for the purpose of evaluating or scoring the suitability of multiple candidate funding assistance programs. They are used to guide generative artificial intelligence models to compare the degree of matching between business content information and the information of each candidate program and output the suitability results.

[0076] "Adaptability calculation" refers to the process by which a server, based on the output of a generative artificial intelligence model or a rule algorithm, quantifies or ranks the degree of matching between each candidate funding assistance program and the business content information.

[0077] The "optimal funding assistance system" refers to the funding assistance system that is determined by the server to be the one that best matches the business content information and the applicable conditions of the system among multiple candidate funding assistance systems, through suitability calculation or comprehensive evaluation.

[0078] "Document style information" refers to template information used to define the structure and presentation of application documents, including the document's chapter composition, field names, field order, and placeholder configurations for filling in business content information.

[0079] "Placeholders" are marks or symbols set in document style information that are to be replaced by specific data content during the document generation process. They are used to indicate the location where specific information such as project name, project summary, and budget content should be filled in.

[0080] "Document generation prompt statements" refer to prompt statements constructed for the purpose of generating or improving the content of application documents. They are used to guide generative artificial intelligence models to generate text or field content that meets the document style requirements based on the parsing results and business content information.

[0081] "Application document data" refers to electronic document data automatically generated based on document style information, parsing results, and business content information, which is used to submit to the recipient of the funding assistance system. This includes structured or semi-structured content such as application forms, project plan descriptions, and budget details.

[0082] "Display data" refers to the visual data that the server converts or packages the application document data for presentation on the user interface, including markup language data, interface component data, or other formatted data suitable for display on user terminals.

[0083] "User interface" refers to the interface components or collection of interfaces that allow users to interact with the system on their user terminals. Through this interface, users can input business information, view application documents, modify content, and submit the application.

[0084] "Additional input information" refers to text or data added by the user through the user interface after the application document is automatically generated, which is used to supplement or refine the application content, such as supplementary explanations, detailed timelines, or risk countermeasures.

[0085] "Modified information" refers to the changes made when users modify the data of generated or displayed application documents through the user interface, including the data corresponding to editing, deleting or rewriting existing fields.

[0086] "Destination information" refers to the address or interface information associated with a specific funding assistance program, used to identify the external processing or receiving device to which the electronic application document should be sent, including network address, interface path, protocol parameters, etc.

[0087] "External processing device" refers to an external information processing device or server system that is connected to this system via a communication network and is used for further processing or reviewing application documents.

[0088] "External processing device" refers to an external information processing device or electronic application system that serves as the main body for receiving and registering application documents under the financial assistance system.

[0089] "Management information" refers to relevant data used to record and track the application processing status within the system, including sending results, acceptance number, timestamp, processing status, and log information related to the application process.

[0090] "User interaction information" refers to behavioral data that can be collected and analyzed by the server, such as input events, click behaviors, dwell time, and editing operation history generated during the process of users interacting with the system through the user interface.

[0091] "Sentiment analysis prompts" refer to prompts designed to analyze the emotional tendencies of users during text input or interaction, and are used to guide generative artificial intelligence models to identify the emotional state of user text content or interaction features.

[0092] "User emotional state" refers to the abstract attribute that represents a user's subjective emotional tendency or psychological state, inferred by performing emotional analysis on user input and user interaction information, such as information categorized as tension, confusion, or positivity.

[0093] "Adjustment information" refers to control parameters or indications generated based on the user's emotional state, used to adjust the style, tone, and level of expression in document generation prompts or application document data.

[0094] "Documentation style" refers to the form and style of information presentation in the application document, including tone, paragraph structure, level of detail in the explanation, and the way terminology is used.

[0095] "Explanation level" refers to the level of detail or abstraction in the explanation of a certain matter in the application document, including the division of levels such as general description, general description, and detailed description.

[0096] "Item-by-item editable area" refers to an independent editing area in the user interface that divides the content of the application document into units of items or fields, allowing users to view and modify each item separately.

[0097] "Changes" refers to the differences that occur after a user modifies the application document data in each editable area, indicating the changes between the original content and the modified content.

[0098] "Append Input Prompt Statements" refer to prompt statements used to supplement or optimize the newly added or changed parts after the user has completed partial content modification, and their content is constructed based on the modified content.

[0099] "Regeneration or completion" refers to the process by which a server uses a generative artificial intelligence model to regenerate, correct, or supplement certain fields or paragraphs of the application document data while maintaining the existing structure and unchanged content.

[0100] In one embodiment of the present invention, a server, a terminal, and a user collaboratively constitute a financial assistance application support system based on a generative artificial intelligence model and dynamic prompt statements. The server improves upon traditional rule-based matching and static template-based computer processing methods by performing a series of data processing and operations, including high-dimensional vectorization of natural language business content, semantic parsing, system matching scoring, and templated document generation. This results in technical improvements in processing speed, matching accuracy, data management structuring, and network communication load.

[0101] In terms of hardware configuration, the server can be a general-purpose server computer, including one or more central processing units (CPUs, such as multi-core general-purpose processors), graphics processing units (GPUs, such as parallel processors for neural network inference), main memory (RAM), non-transitory storage devices (hard disks or solid-state drives), and network interface devices (Ethernet interfaces or wireless network interfaces). The server can be deployed in data centers or cloud computing environments. The terminal can be an information processing device with a display and input device, such as a laptop, desktop computer, tablet, or smartphone, running a general-purpose operating system (such as a Unix-like operating system, a general-purpose desktop operating system, or a mobile operating system) and a web browser or dedicated client application. Users interact with the server through a user interface on the terminal.

[0102] In terms of software architecture, a server may include: a network service module (e.g., a web server based on the HTTP / HTTPS protocol), an application logic module (e.g., business logic implemented using a general server-side framework), a database management module (e.g., a relational database management system), a generative AI invocation module (for data interaction with generative AI models), a template management module (for managing the template structure of application documents), a document generation module (for generating document data based on templates and parsing results), and a log and management information module. The database management module can use relational database software (e.g., a database system that supports SQL) to store information on funding assistance programs, application input information, application document data, and log information in a table structure. The template management module can manage template files in HTML, XML, or DOCX formats.

[0103] The server uses a generative AI model in the generative AI invocation module. This generative AI model can be a neural network language model based on a multi-layer Transformer structure. The Transformer structure includes multiple self-attention sublayers and feedforward neural network sublayers, each layer performing multi-head self-attention operations and residual connections and layer normalization operations. The model's parameters include word embedding matrices, positional encoding parameters, self-attention weight matrices (query, key, value matrices), and the weights and biases of the feedforward network. During the training phase, the model updates the weights by maximizing the conditional probabilities of the training corpus or minimizing the cross-entropy loss function. Gradient descent with Adam or similar variants can be used during training, with gradients calculated and model parameters updated via backpropagation. Training data can include general natural language corpora as well as domain corpora related to policy descriptions, funding aid system descriptions, and project applications. Data augmentation techniques, such as synonym replacement, sentence shuffling, and masked fill-in, can be used during training to improve the model's robustness to diverse expressions.

[0104] During the inference phase, the server performs the following technical processing on the natural language business content input by the user. After the terminal transmits the user-input text to the server, the server uses an embedding layer to map each token into a vector representation and uses a multi-head self-attention mechanism to calculate the correlation between different words within the text, thereby obtaining the contextual representation of the text in a high-dimensional space. The server constructs parsing prompt statements in the application logic module, combining the "system role instructions" and the "user-provided business content text" into the generative artificial intelligence model. An example of a parsing prompt statement is as follows: Example of a system prompt statement: "You are a financial assistance system analysis assistant. Please extract information such as keywords, project type, industry sector, project goals, target audience, and implementation area from the project description and output it in a structured format." Example of a user prompt statement: "Project content: We plan to leverage local tourism resources to host summer events, including hot spring experiences, specialty markets, and concerts, to attract tourists from other areas and increase the region's visibility." The server parses the business classification information, target domain information, target subject information, and regional information from the text output by the model into structured application input information. To achieve this, the server defines a set of field mapping rules and regular expression parsing rules in the application logic module to extract key-value pairs from the model's output text. For example, the server can require the model output to contain obvious tags, such as "Project Type:", "Industry Domain:", "Target Audience:", "Region:", etc. The application logic module performs string splitting and field filling based on these tags. Compared to simple keyword search, this structured parsing method allows the server to form a unified data structure internally, facilitating efficient execution of subsequent retrieval and scoring algorithms, thereby improving overall processing efficiency.

[0105] The server accesses the financial assistance program information table using the database management module. This table can have fields such as: Program ID, Applicable Industry, Applicable Region, Target Entity Type, Program Category, Subsidy Limit, Application Period, and Text Description. The server maps the parsed business classification information, target industry information, target entity information, and regional information to database query conditions, and performs filtering through an index structure (e.g., a B+ tree index) to quickly retrieve several candidate financial assistance program records. To further improve matching accuracy, the server can construct evaluation prompts, inputting the business content and candidate program description text into a generative artificial intelligence model, allowing the model to output a suitability score and reason for each program. For example: Example of an evaluation prompt statement: "Below is a project description and explanations of several funding assistance programs. Please give each program a matching score from 0 to 100, and briefly explain the reasoning behind the score."

[0106] Project Description: We plan to leverage local tourism resources to host summer events, including hot spring experiences, specialty markets, and concerts, to attract tourists from other areas and increase the region's visibility.

[0107] Explanation of Financial Assistance System A: ... Explanation of Financial Assistance System B: ... Explanation of Financial Assistance System C: ..." After receiving the output, the server parses the score values ​​into floating-point data and selects the system with the highest score as the optimal funding assistance system using a ranking algorithm. In this way, the server not only relies on simple label matching but also utilizes high-dimensional semantic similarity for system selection, thus achieving higher matching accuracy than traditional rule-based systems. By caching intermediate embedding vectors or reusing existing parsing results for similar business content, the server can reduce redundant calculations and improve inference throughput.

[0108] The server uses a template management module and a document generation module working together for document generation. The server selects document style information corresponding to the optimal funding assistance system from the template library, such as an application template for a specific system. This template is a structured resource, and the field positions and display order can be defined using HTML or markup language. Placeholders are preset in the template, such as "{{Project Name}}", "{{Project Purpose}}", "{{Implementation Content}}", and "{{Budget Summary}}". The server maps the parsed results to the original business content, filling each field into the corresponding placeholders to generate preliminary application document data. For sections requiring the model to generate complete paragraphs, the server constructs document generation prompts, such as: Example of document generation prompt statement: "You are an expert in writing financial aid applications. Based on the following project content, please generate a summary section suitable for submission to the financial aid system, including project background and purpose, implementation details, and expected outcomes:" Project Content: We plan to leverage local tourism resources to host summer events, including hot spring experiences, specialty markets, and concerts, to attract tourists from outside the area and increase regional awareness. The server inputs the prompt statement into a generative AI model, obtains continuous text output, and then fills the corresponding paragraph area in the template with this text. Because the server controls the content of the prompt statement and the template structure, the model output is constrained within predetermined fields and formats, avoiding unstructured results and simplifying subsequent data management. Compared to the traditional method of generating an entire free-form document at once, this implementation achieves higher editability and traceability through field-level prompt statement control.

[0109] In terms of user interface, the terminal renders the application document data returned by the server as editable areas for each item. For example, the terminal displays the "Project Name" field as a single-line text box, the "Project Purpose and Background" field as a multi-line editable area, and the "Budget Details" field as an editable table. Users can modify or append to the content of each field on the terminal. For example, users can add detailed numerical targets to the "Expected Results" column. The terminal sends the user's modifications as differential data to the server.

[0110] After receiving the differential data, the server does not regenerate the entire document. Instead, it constructs append input prompts only for the changed fields, and calls a generative AI model to generate relevant supplementary content or perform semantic checks and style adjustments. For example: Example of a prompt statement for appending input: "The following is a modified description of the project purpose. Please adjust the wording to make it clearer and more in line with common expressions used in financial aid applications, while maintaining the original meaning:" The user-edited text: ..." The server then regenerates or completes local fields accordingly. This local call method reduces the amount of data transmitted and computation, lowers the communication load between the server and the terminal, and improves response speed by reducing unnecessary overall rewriting. This control method based on field differentiation and appending prompts differs from manual editing by humans and represents a special optimization of the computer's internal data structure and call flow.

[0111] In another implementation, the server also uses a generative artificial intelligence model to infer the user's emotional state, such as tension, confusion, or confidence, based on user interaction information and sentiment analysis prompts. The server can construct sentiment analysis prompts, for example: Example of sentiment analysis prompts: "You are a sentiment analysis assistant. Please determine the user's current emotional state (e.g., nervous, confused, positive, etc.) based on the user's input text below, and provide a reason:" User input text: "I'm not very familiar with this application process, and I don't understand many of the technical terms. I'm worried I'll make a mistake." The server uses the sentiment category output by the model as input to adjust the tone and explanatory level of subsequent prompts. For example, when a user is judged to be confused, the server adds constraints such as "please use more understandable expressions and add appropriate explanatory sentences" when generating prompts for the document, making the explanatory section of the generated application document more detailed. The server can also add guidance explanations to the user interface prompt text. This technical processing of dynamically modifying prompts based on sentiment states allows the server to control the generative AI model to output at different levels of explanation, thereby adjusting the level of detail in the calculated output without changing the underlying business logic, improving readability, reducing the number of times users need to modify the prompts, and indirectly shortening the overall session length and server processing time.

[0112] In terms of internal model processing, the server can generate semantic vector embeddings for business content text and policy description text, calculate the matching degree in the vector space using cosine similarity or other metric algorithms, and combine this value with the score output by the generative artificial intelligence model to reduce the impact of model output noise. The server can use weighted linear combination or learned regression models to fuse multiple scoring sources, thereby technically improving the stability and robustness of the matching results.

[0113] In terms of data structure, the server constructs a unified data object for each application, including: the original business content text, parsed result fields, a list of candidate policies and their scores, the selected policy ID, application document field content, version number, user modification differential records, and sending logs. By maintaining unique identifiers and version tags for each data object, the server achieves traceable state migration in the database management module, enabling rapid recovery of any version of the data state during subsequent dialogues or corrections. This structured data management, compared to the traditional method of only storing the final document, improves the server's recovery speed and consistency maintenance capabilities in multi-round interaction scenarios.

[0114] In another variant implementation, the server can deploy the model inference engine locally instead of using external cloud-based generative AI services. The server leverages GPUs to perform matrix multiplication and attention calculations, significantly increasing the number of applications that can be processed per unit time by batch processing multiple user requests. The server can pre-compute vectors of frequently occurring policy explanatory texts and cache them in memory. Once the parsing results are determined, only the similarity of the business content vectors needs to be calculated, eliminating the need to repeatedly input the policy explanatory texts into the model each time, thereby reducing computational costs and latency.

[0115] Through the aforementioned structural design and algorithm arrangement, the server does not simply replace human form filling with automatic filling. Instead, it utilizes generative artificial intelligence models, multi-layered Transformer structures, dynamic prompt statement control, field-level differential updates, and structured data management to construct a highly efficient semantic processing pipeline for natural language business content. Within this pipeline, through meticulous data structure design and call sequence control, the server achieves the following technical effects: improved accuracy of automatic conversion from natural language to structured data; improved accuracy and stability of matching financial assistance systems; faster response speed in the application document generation and modification process; reduced network communication data volume; improved server resource utilization; and reduced error rate and frequency of human intervention. Therefore, at the computer technology level, this invention improves the architecture, information retrieval and matching algorithms, document generation mechanisms, and human-computer interaction control methods of natural language processing systems.

[0116] use Figure 11 The processing flow is explained.

[0117] Step 1: Users input business information using a terminal and send it to the server.

[0118] Users input business information such as project name, project purpose, implementation details, budget, and implementation region in natural language text format via keyboard or touch keyboard on the terminal, or in input forms within a browser interface or dedicated application. The terminal takes the text from each input field as input, assembles it into a data structure containing multiple key-value pairs, and serializes it into a JSON string. The terminal performs data processing operations based on this JSON string: adding metadata such as timestamps, user identifiers, and session identifiers to form a complete request message. Using this message as input, the terminal sends a POST request to the server's predefined API address via the HTTPS protocol. The input for this step is the natural language text entered by the user in the interface and the basic form fields; the output is a network request message containing business content information and metadata.

[0119] Step 2: The server receives and parses the service content requests sent by the terminal.

[0120] The server takes the HTTP request message from the terminal as input. First, the network service module decrypts the TLS layer data and parses the HTTP header and request body. Based on the request path and method, the server distributes the request to the corresponding processing function in the application logic module. The application logic module calls a JSON parsing library to parse the JSON string in the request body into an internal data structure (such as a dictionary or object). The server checks if required fields exist and if the text length is within a preset range, returning an error response for invalid data. In data processing, the server maps the parsed fields to an internal unified data structure and writes the original business content text along with the user ID and timestamp into the "Application Input Table" in the data storage device. The input for this step is the JSON request message sent by the terminal, and the output is a structured application input information record and a receipt confirmation response available internally to the server.

[0121] Step 3: The server constructs and parses the prompt statement and calls a generative artificial intelligence model to perform semantic parsing.

[0122] The server takes the stored business content text as input, reads the text from the database or cache, and constructs parsing prompts in the application logic module. The server generates system prompts, such as: "You are a financial assistance system parsing assistant. Please extract keywords, project type, industry sector, project goals, target audience, implementation area, etc. from the project description and output them in a structured manner." It also generates user prompts, such as: "Project content: ... (original business content)...". The server combines the system and user prompts into a model input sequence, which is then used as input to the generative AI model's calling module. After receiving this combined text, the generative AI model internally performs a series of matrix operations, including word embedding, positional encoding, multi-head self-attention, and feedforward networks, to vectorize and model the text within its context. Based on pre-trained weights, the model generates output text token by token, containing tagged structures such as "Project type:...", "Industry sector:...", "Target audience:...". The server receives the model's output text, calls string parsing and regular expression matching functions, and extracts the content after these tags into fields such as business classification information, target sector information, target subject information, and region information. The input for this step is the original business content text and the parsing prompt statement, and the output is a parsing result data structure containing multiple fields.

[0123] Step 4: The server converts the parsed results into database search criteria and extracts candidate funding assistance programs.

[0124] The server takes the parsing results obtained in step 3 as input and reads the project type, industry sector, target subject category, and regional information. In the application logic module, the server maps these fields to database query conditions; for example, it maps "tourism promotion" to the industry field value "tourism industry," and the region field to a specific administrative region code. The server constructs an SQL query statement or generates a query expression using ORM, and performs retrieval operations on the funding assistance program information table using the database management module. The database searches the fields based on indexes (such as B+ tree indexes) and returns a set of funding assistance program records that meet the conditions. The server performs preliminary filtering on the returned records, such as removing programs that have exceeded their application deadline or whose budget limits are clearly incompatible. The server saves the filtered program records as a list of candidate funding assistance programs. The input to this step is the parsing result data structure and the funding assistance program information table; the output is a list of candidate programs with multiple candidate program entries.

[0125] Step 5: The server constructs evaluation prompts and calculates the suitability of candidate systems.

[0126] The server takes a list of candidate policies and the original business content text as input, and generates a set of evaluation prompts for each candidate policy. The server can construct comprehensive evaluation prompts, such as: "Below is a project description and several financial assistance policy descriptions. Please give each policy a matching score from 0 to 100, and briefly explain the reasoning behind the score. Project description: ... Financial assistance policy A description: ... Financial assistance policy B description: ...". The server concatenates this prompt with the business content text and policy description text as input, and calls a generative artificial intelligence model. Internally, the model performs joint encoding on the project description and each policy text, calculates semantic relevance, and outputs text results similar to "Policy A: 85 points, reason: ...; Policy B: 60 points, reason: ...". The server parses this output, extracts the score for each policy as a numerical value, and maps it to the corresponding record in the candidate policy list. The server can further calculate the cosine similarity between the business text and policy description text based on vector similarity methods, and weight this value with the model's output score to obtain the comprehensive fit. The server uses a sorting algorithm to rank all candidate systems from highest to lowest suitability, selecting one or more systems with the highest suitability as the optimal funding assistance system. The input for this step is a list of candidate systems, business content text, and evaluation prompts. The output is a sorted list with suitability scores and at least one system record marked as the optimal funding assistance system.

[0127] Step 6: The server selects a document template and maps the parsed results to the document structure.

[0128] The server takes the optimal funding assistance system record and parsing result data structure as input and retrieves the document style information corresponding to the system from the template management module. The document style information includes field layout, chapter structure, and placeholder names, such as "{{Project Name}}", "{{Project Background}}", "{{Implementation Content}}", and "{{Budget Summary}}". Based on predefined mapping rules, the server maps the project type, industry sector, target entity, and regional information from the parsing results, as well as the project name and summary from the original business content, to the corresponding placeholders in the template. The server constructs a document data structure in memory, where each placeholder corresponds to a field value. For fields that do not require large blocks of text to be generated by the model, the server directly fills them with existing content. The inputs to this step are the optimal funding assistance system information, parsing results, and document template; the output is an intermediate document data structure with placeholder values.

[0129] Step 7: The server constructs a document, generates prompt statements, and generates a text paragraph filling template.

[0130] The server takes the unfilled fields in the intermediate document data structure and the original business content text as input, and generates document generation prompts for fields requiring detailed description (such as "Project Background and Purpose," "Implementation Methods," and "Expected Results"). For example, the server generates the prompt: "You are an expert in writing financial aid applications. Please generate a summary section of the application suitable for submission to the financial aid system based on the following project content, including project background and purpose, implementation content, and expected results: Project content: ...". The server inputs this prompt into a generative artificial intelligence model, which internally uses a Transformer decoder structure to generate natural language paragraph text word by word based on context vectors. The server receives the model output, truncates or normalizes the generated text (such as removing redundant introductions and standardizing punctuation), and then fills it into the corresponding fields in the document data structure. The server repeats this process to generate content for multiple paragraph fields. The inputs for this step are the intermediate document data structure, business content text, and document generation prompts; the output is application document data with all placeholders filled.

[0131] Step 8: The server converts the application document data into display data and sends it to the terminal.

[0132] The server takes the complete application document data as input, calls the document generation module, and renders the field content into a structure suitable for front-end display, such as HTML fragments or structured JSON (divided by chapters and fields). Following a predetermined layout, the server marks each field with editable attributes (read-only or editable), field type (single-line, multi-line, table), and validation rule hints. The server packages the rendered result into a response message and sends it to the terminal via HTTPS. The terminal receives this response, uses the display data within as input, and utilizes the front-end framework to display each field as an editable area for the user to view and modify. The input for this step is the application document data, and the output is the display data sent to the terminal and the editable interface rendered locally on the terminal.

[0133] Step 9: Users can review application documents and make modifications or add input on the terminal.

[0134] Users take the application document interface returned by the server as input and read through the fields on the terminal screen, including project name, background and purpose, implementation content, budget, and expected results. Users can edit certain fields on the terminal interface according to the actual situation, such as modifying the project name description, adding budget details, and adding time plans. The terminal tracks the original and current values ​​of each field locally, generates a differential data structure for the changed parts, and records the changed field identifiers and the modified text content. The terminal can send this differential data as output to the server via HTTPS each time the user finishes editing or clicks "Save Draft". The input for this step is the application document interface and user operations; the output is a differential update message containing the changed fields and their new values.

[0135] Step 10: The server receives user modifications and performs partial regeneration or completion.

[0136] The server takes the differential update message sent by the terminal as input, parses each changed field, writes the new text content into the corresponding position in the application document data structure, and updates the current version of the field in the database. For fields that require language optimization or content completion by a generative AI model, the server constructs an append input prompt, such as: "The following is the user's modified description of the project purpose. Please adjust the sentence to make it clearer and more in line with the common expressions in funding assistance applications while maintaining the original meaning: User's modified text: ...". The server calls the generative AI model with this prompt and the user's modified text as input, and the model outputs the polished or supplemented text. After confirming that the output content does not destroy key elements through string comparison and rule validation, the server overwrites or merges it into the corresponding field. The server only performs this regeneration process on fields marked as changed, without regenerating the entire document, thereby reducing computation and communication load. The input for this step is the differential update message and the append input prompt, and the output is the updated application document data and the updated database record.

[0137] Step 11: The user confirms the final document in the terminal and issues a submit command.

[0138] The user takes the latest application document returned by the server as input and double-checks all fields for accuracy, including the selected funding assistance program name, program requirements, application period, and descriptions. After confirming everything is correct, the user clicks the "Submit Application" button on the terminal interface. The terminal performs form validation locally, checking that all required fields are filled in and formatted correctly. It then encapsulates the application ID, current document version number, and submission command into a submission request message. The terminal sends this message to the server via HTTPS. The input for this step is the final document content displayed on the terminal and the user's actions; the output is a network request message containing the submission command.

[0139] Step 12: The server completes the application submission process and records management information.

[0140] The server takes the submission request message as input, reads the latest application document data and status information from the database, confirms that the application is in a submittable state and has not passed the deadline. The server updates the application status to "submitted," generates or updates the acceptance number, and writes the submission time into the management information table. Based on the destination information of the specific funding assistance program, the server constructs a transmission message, which may contain PDF format application documents or structured data, and sends it to an external processing device or external acceptance device through a predetermined interface (such as an external API or email gateway). The server receives the acknowledgment information (such as the acceptance number or processing status) returned by the external device and writes it, along with the transmission result, into the management information table. Finally, the server returns a submission success response to the terminal, including information such as the acceptance number and submission time. The inputs to this step are the submission request message and the current application document data in the database; the outputs are the updated management information record and the submission result response sent to the terminal.

[0141] Application Example 1 The process flow corresponding to the specific processing in Use Case 1 will be described below. The various parts of the system described below are implemented by the data processing device 12 and the intelligent device 14. Furthermore, the data processing device 12 is referred to as the "server" and the intelligent device 14 is referred to as the "terminal".

[0142] Existing computer-based financial aid application support technologies typically rely on rule-based matching or simple keyword retrieval to search for financial aid programs corresponding to user-inputted project descriptions from storage. This approach suffers from several limitations at the computer technology level: First, the server's ability to parse natural language text is limited, hindering deep semantic structuring of the content at the system level. This results in coarse search criteria, low accuracy, and the server's need to traverse numerous irrelevant records during database retrieval, increasing storage access load and response latency. Second, the automatic generation of application documents largely depends on fixed templates and static replacements. The server lacks the ability to automatically generate and dynamically adjust explanatory text, requiring users to perform extensive manual editing on the terminal, increasing the number of human-computer interaction rounds and network requests, negatively impacting overall system throughput and interactive performance. Third, existing systems often loosely connect natural language processing, database retrieval, and document generation, failing to establish a unified structured data flow and a prompt-driven generative AI model invocation mechanism on the server side. This leads to redundant data conversion between modules, increased memory consumption, and difficulties in scaling and maintaining the processing flow.

[0143] Therefore, a new computer implementation is needed to integrate natural language parsing, structured data generation, precise matching of financial assistance systems, generation of initial drafts of application documents, and automatic rewriting of explanatory texts in a single system on the server. This would improve server processing efficiency, storage access patterns, and human-computer interaction processes at the system architecture level, thereby enhancing the overall computing performance and scalability of the financial assistance application support system.

[0144] The specific processing performed by the specific processing unit 290 of the data processing apparatus 12 in Application Example 1 is achieved by the following means.

[0145] In this invention, the server includes: a module for receiving text information containing activity content via a communication network through a user terminal and storing the text information in a storage device; a module for generating prompt statements for a generative artificial intelligence model, inputting the text information into the natural language parsing function of the generative artificial intelligence model, extracting business domain, activity type, scale information, and multiple feature words from the text information through parsing processing, and generating structured data; a module for searching an information storage device containing registered financial assistance system information based on the structured data, calculating evaluation values ​​for multiple candidate financial assistance systems based on industry fit, activity fit, and condition matching, and selecting the optimal financial assistance system based on the evaluation values; and a module for obtaining a document template based on the format information contained in the application requirements of the selected financial assistance system, and using the structured data and the text information to process the application template. The application document template includes a module for automatically assigning values ​​to predetermined items to generate an initial version of the application document data; a module for generating prompt statements for a generative artificial intelligence model for at least a portion of the purpose description, background description, or effect description in the initial version of the application document data, automatically generating or modifying explanatory text through the text generation processing of the generative artificial intelligence model, and reflecting the explanatory text in the initial version of the application document data; a module for sending the initial version of the application document data and the reasons for recommendation related to the selected funding assistance system to the user terminal through a user interface, and reflecting the corrections received from the user terminal in the initial version of the application document data to generate a final version of the application document data; and a module for converting the final version of the application document data into a predetermined electronic document format or predetermined data structure and sending it to an external review device to complete the application processing. This allows for the formation of a data flow centered on structured data within the server, enabling the controllable invocation of the parsing and generation capabilities of generative artificial intelligence models through prompts. This achieves high-precision semantic structuring of natural language input, efficient and precise matching of financial assistance systems, and automatic generation and dynamic rewriting of application documents. Consequently, it reduces invalid database access and repeated template rendering, lowers server computing and storage resource consumption, shortens the number of human-computer interaction rounds between user terminals and servers, and improves the overall processing performance and scalability of the financial assistance application support system.

[0146] "System" refers to an overall device and its collaborative operating environment consisting of at least one server, at least one user terminal, and a communication network connecting the two, used to perform functions such as information reception, data processing, document generation, and data transmission.

[0147] A "server" is a computing device equipped with a processor, storage device, and communication interface, used to execute program instructions to perform natural language parsing, database retrieval, document generation, and data interaction with user terminals and external review devices.

[0148] "User terminal" refers to a computing device operated by a user and used to send and receive data with a server through a communication network, including but not limited to smartphones, tablet computers, personal computers or other electronic devices with user interface functions.

[0149] "Communication network" refers to wired or wireless networks used to transmit data between servers and user terminals, and between servers and external review devices, including but not limited to local area networks, wide area networks, mobile communication networks, and the Internet.

[0150] “Text information” refers to character data recorded in natural language form that is entered by the user on the user terminal and sent to the server, including activity content, project descriptions, and explanatory text related to financial assistance.

[0151] "Activity content" refers to a summary description of the business activities or projects that the user plans to implement, including a natural language expression of information such as the business area, implementation purpose, implementation method, and expected results.

[0152] "Storage device" refers to storage resources used in servers to store programs, text information, structured data, financial assistance system information, and application document data, including main memory and secondary memory.

[0153] "Generative artificial intelligence models" refer to data processing models trained based on machine learning and deep learning techniques that can automatically generate or parse natural language text based on input prompts, including but not limited to models used for text understanding and text generation.

[0154] "Prompt statements" refer to natural language instructions generated by the server and input into the generative artificial intelligence model, which specify the content and style of the parsing task, text generation task, or rewriting task to be performed by the model.

[0155] "Natural Language Parsing Function" refers to the functional module or processing procedure in a generative artificial intelligence model that performs word segmentation, semantic understanding, entity recognition, keyword extraction and classification on the input text in order to output structured results.

[0156] "Business domain" refers to the industry category or business type classification information that is extracted from text information through natural language processing and is related to user activity content.

[0157] "Activity type" refers to information extracted from text information through natural language processing, which represents the category attribute of the activity or project that the user plans to implement.

[0158] "Scale information" refers to data extracted from text information through natural language processing, used to represent the size of an activity or project in terms of funding, personnel, time, or organizational scale.

[0159] "Feature words" refer to keywords or important terms extracted from text information through natural language processing that can reflect the key points of the activity or are related to the conditions for financial assistance.

[0160] "Structured data" refers to a collection of data that is organized based on the results of natural language parsing, and represented in the form of predefined fields or data structures, which is conducive to database retrieval and logical operations.

[0161] "Information storage device" refers to storage resources that store information on funding assistance systems, document templates, and parameters related to the system. It can consist of one or more database systems or file storage systems.

[0162] "Information on financial assistance programs" refers to data records regarding the conditions, eligible recipients, subsidy content, application requirements, time limits, and other relevant rules for various financial assistance programs.

[0163] "Candidate funding assistance schemes" refers to various funding assistance schemes in a set of potentially applicable funding assistance schemes retrieved from structured data in an information storage device.

[0164] "Industry fit" refers to an evaluation metric used to represent the degree of matching between the applicable industries of a candidate funding assistance program and the business areas obtained through natural language parsing.

[0165] "Activity fit" is an evaluation metric used to represent the degree of matching between the types of support activities of a candidate funding assistance program and the types of activities obtained through natural language parsing.

[0166] "Condition matching degree" refers to the evaluation index used to represent the degree of matching between the application conditions of a candidate funding assistance system and the conditions such as scale information and feature words contained in structured data.

[0167] "Evaluation value" refers to a numerical value or score obtained by using predetermined calculation rules based on one or more indicators such as industry fit, activity fit, and condition matching to comprehensively evaluate the fit of candidate funding assistance systems.

[0168] The "optimal funding assistance system" refers to the funding assistance system selected from multiple candidate funding assistance systems based on the evaluation value, which has the highest degree of fit with the overall user activity content.

[0169] "Document template" refers to a predefined document structure or format data used to generate application documents, which contains multiple predefined items and their placeholders for receiving automatically assigned content.

[0170] "Pre-set items" refers to fields or columns pre-set in the document template for filling in specific information, including multiple information items such as project name, purpose description, and budget amount.

[0171] "Initial version data of application documents" refers to the application document data generated by the server based on the document template, using structured data and text information to automatically assign values ​​to predetermined items, before the user makes any corrections.

[0172] "Purpose statement" refers to the text in the application documents that explains the purpose and intent of the user's planned activities or projects.

[0173] "Background information" refers to the textual content in the application documents that explains the background, current problems, and necessity of implementing the activity or project.

[0174] "Explanation of effects" refers to the textual content in the application documents used to explain the expected effects or results after the implementation of the activity or project.

[0175] "Explanatory text" refers to natural language text content generated or modified by a generative artificial intelligence model based on prompts, used as a purpose description, background information, or effect description.

[0176] "Reasons for Recommendation" refers to the explanatory text used to explain to users the compatibility between the optimal financial assistance system and the user's activity content, including explanations of applicable conditions, selection criteria, or advantages.

[0177] "Final version application document data" refers to the application document data used for formal submission, which is obtained after reflecting the revised content received from the user terminal and the explanatory text generated or modified by the generative artificial intelligence model, based on the initial version of the application document data.

[0178] "Electronic document format" refers to a predefined file format used to store and transmit the final version of application document data, including text files, portable document format files, or other electronic document file formats.

[0179] "Data structure" refers to the logical structure or encoding form used to organize and represent the final version of application document data, including key-value pair structure, hierarchical structure, or markup language structure.

[0180] "External review device" refers to a computing device or information processing system that is connected to a server via a communication network and is used to receive the final version of application documents and to process and review applications for financial assistance.

[0181] "User interface" refers to the interface between the server and the user terminal used for interactive display and data input / output, including the graphical user interface generation module and the communication interface used to transmit interface data.

[0182] In one embodiment of the present invention, a server serves as the core computing device, comprising a processor, main memory, auxiliary memory, a network interface, and a storage system connected thereto. The server runs application server software (e.g., a general-purpose web server and application runtime environment) on an operating system (e.g., a Linux-based server operating system), and further runs applications used to implement the functions of the present invention. A terminal serves as the user-side computing device, comprising a processor, storage device, display device, and input device (touchscreen, keyboard, pointing device, etc.). The terminal runs browser software or dedicated applications on a general-purpose operating system (e.g., a mobile operating system or a desktop operating system) to conduct bidirectional data communication with the server via a communication network. The user interacts with the server through the terminal's graphical interface.

[0183] The server stores programs in auxiliary storage for implementing the system functions of this invention, including: a communication management module, a natural language parsing module, a generative artificial intelligence invocation module, a structured data generation module, a funding assistance system matching module, a document template management module, an application document generation module, a verification and error checking module, an external review device interface module, and a user interface management module. The server also stores various data structures in the database management system, including: a user information table, a raw text information table, a funding assistance system table, a document template table, an application document table, and a log table. The database management system can be implemented using relational database management software, such as a general-purpose relational database system.

[0184] When performing natural language parsing and text generation, the server employs a generative artificial intelligence model. In one implementation, this model can be a neural network language model based on a Transformer architecture. During model deployment, the server stores the trained model parameter file in high-speed storage and loads it into memory during inference. The model comprises a multi-layer self-attention encoder and decoder structure, with each layer including a multi-head self-attention sublayer and a feedforward fully connected sublayer. Model parameters include word embedding matrices, positional encoding parameters, multi-head attention weight matrices, feedforward network weight matrices, and bias terms. During natural language parsing, the server uses the model's encoding part to encode the input text information into a high-dimensional semantic vector representation; during text generation, the server uses the model's decoding part to autoregressively output the target text sequence given prompts.

[0185] When training a generative AI model, the server uses multi-domain text corpora stored in the training data storage device, including project descriptions, policy documents, and application sample documents. During the training phase, the server employs supervised learning, defining a cross-entropy loss function as the error function, calculating gradients through backpropagation, and updating model weights using stochastic gradient descent optimization algorithms (such as Adam). The server uses batch training, combining multiple samples into small batches and utilizing vectorized operations for parallel computation on accelerated hardware (such as general-purpose graphics processing units) to improve training speed. The server can increase the diversity of training samples through data augmentation techniques (such as synonym replacement, sentence structure transformation, and slight noise injection), thereby improving the model's robustness and generalization ability to different expressions.

[0186] When the server invokes the generative AI model during the inference phase, it no longer updates the weights but instead performs forward propagation operations using the fixed model weights from the training. In the natural language parsing task, the server converts the user-input text information into a token sequence. The server calculates the context vector representation of each token through an embedding layer and a multi-layer self-attention encoder, and at the top level, maps the vector to category labels (such as business domain, activity type) or keyword probability distributions using a classification head or linear mapping + activation function. During keyword extraction, the server employs thresholding or Top-K selection mechanisms to select multiple feature words from the probability distribution. In scale information extraction, the server combines regular expression matching with model output results to structure numerical information such as amount, number of personnel, and implementation period.

[0187] In the structured data generation module, the server organizes the output of the natural language parsing module into a unified data structure, such as record objects containing fields like "industry classification," "project type," "scale parameters," "keyword list," and "regional information." The server stores these record objects in a structured data table in the database and assigns a unique identifier to each record for subsequent retrieval and association. By using a unified data structure, the server can avoid repeatedly parsing the original text when calling the funding assistance system matching module and document generation module multiple times, thereby reducing computational overhead and latency.

[0188] In the funding assistance system matching module, the server accesses the funding assistance system table, which stores multi-dimensional condition fields for each system in a relational database format. These fields include "Applicable Industry Classification," "Supported Project Type," "Minimum and Maximum Applicant Size," "Applicable Region," "Available Budget Range," and "Keyword Tag Set." The server pre-calculates and stores vectorized features for each funding assistance system, encoding industry classification, project type, and keyword tags into multi-dimensional sparse or dense vectors. During matching, the server converts the industry classification, project type, and keyword list from the structured data into query vectors using the same encoding rules. The server calculates similarity indices, such as cosine similarity or weighted Hamming similarity, between the query vector and each funding assistance system vector, and performs range judgments based on numerical conditions (such as budget range and size range). The server calculates evaluation values ​​using a weighted scoring function, which can be a multinomial linear weighted model or a piecewise function with a threshold, thus obtaining a comprehensive fit score for each candidate funding assistance system. The server sorts the candidate systems according to their scores and selects the one or more with the highest scores as the optimal funding assistance system.

[0189] When generating recommendation reasons, the server utilizes matching process information recorded in structured data (such as which conditions match, which fields are ignored, and the composition of the matching score) to construct prompt statements, embedding this structured matching information in natural language. For example, the server generates the following prompt statement: "Please generate a brief recommendation based on the following criteria, explaining why this funding assistance program is suitable for the user: Industry = Retail, Project Type = Cashless Payment System Implementation, Business Size = Small and Medium-sized Enterprises, Matching criteria include: Industry Fully Matched, Project Type Fully Matched, Budget Range Covers User Needs." The server inputs the prompt statement into the generative AI model. During the decoding phase, the server constrains the output length and tone (e.g., by setting temperature parameters and maximum generation length) to generate a coherent recommendation text. Because the server explicitly provides matching elements in the prompt statement, the model-generated description remains consistent with the internal scoring logic, resulting in interpretable recommendations. This prompt-driven generation mechanism, linked to the internal data structure, achieves diverse and highly readable description outputs without adding a large number of hard-coded rules, compared to simple rule templates.

[0190] In the document template management module, the server stores various application document templates in a database or file system. Each template defines multiple predefined items and corresponding placeholders using a template language, such as "Project Name," "Project Background," "Implementation Content," "Total Budget," and "Contact Information." When generating a document, the server loads the corresponding template from the template storage based on the "Template ID" field of the selected funding assistance system and constructs an internal representation (e.g., in the form of an abstract syntax tree or a field list). The server uses structured data and raw text information to perform mapping and assignment on the structured fields in the template; for example, mapping "Industry Classification" to the "Business Area" field in the template, and mapping the parsed monetary information to the "Total Budget" field.

[0191] When generating the purpose description, background description, or effect description, the server constructs prompts for the generative artificial intelligence model, such as: "Based on the following project summary, please generate a project objective statement suitable as a funding application, approximately 300 words in length, using a formal tone: 'We plan to introduce a new cashless payment system in local stores to improve checkout efficiency and integrate with multiple payment platforms.'" Please revise the following project objective into a more formal, logically clear, and suitable wording for a funding application: 'Improve checkout speed, reduce queues, and attract younger customers.' The server combines these prompts with necessary contextual information (such as project type, industry, and scale) as model input, and then uses a generative AI model to decode and output explanatory text. Because the server explicitly limits the length, tone, and purpose of the prompts, the model output text is more suitable in style and structure for direct embedding in application documents, reducing the amount of manual editing required by users on the terminal, thereby reducing the number of human-computer interaction rounds and network communication load.

[0192] After receiving the initial version of the application documents and the reasons for recommendation from the server, the terminal parses the data structure in its local application and presents it in a graphical interface. The terminal uses input components to display the current values ​​of each predefined item, while also marking fields generated or automatically populated by a generative artificial intelligence model. Users can modify, supplement, or delete specific fields on the terminal. The terminal can perform local format checks (such as character limit prompts) during user input to improve the user experience. After user confirmation, the terminal resends the corrected field values ​​to the server.

[0193] In the verification and error checking module, the server performs system-level validation on the application document data uploaded by the terminal. The server checks the existence of required fields, performs range and format checks on numeric fields, and performs regular expression matching and logical checks on date fields (e.g., start date earlier than end date). When incompleteness or errors are detected, the server generates an instruction message containing the error location, error type, and correction suggestions, and sends it to the terminal. This allows the terminal to highlight the erroneous field on the interface and guide the user to correct it. Because error checking is uniformly implemented internally by the server, the server can centrally update rules and thresholds without repeatedly implementing complex validation logic on multiple terminal sides, thereby reducing the burden on the terminals and ensuring consistency.

[0194] After the final version of the application document data is generated, the server converts the data into a predetermined electronic document format or predetermined data structure within the interface module of the external examination device. For example, the server renders the application content into a portable document format file using a document generation library, or encodes structured data into a markup language format for transmission via the interface. The server can compress and encrypt the data before transmission to reduce communication load and improve security. Upon receiving the data, the external examination device can directly parse these standard format data, enabling automatic data entry and workflow distribution, thereby reducing manual data entry.

[0195] In terms of technical effectiveness, the server, through the aforementioned structured data generation and vectorized matching mechanisms, significantly reduces the scanning of irrelevant records in the database, lowers disk access frequency and index traversal length compared to traditional systems that only use keyword string matching, thereby accelerating retrieval speed. The server avoids repetitive natural language parsing operations by using unified structured data objects, improving CPU and accelerator utilization. The server generates explanatory text and recommendation reasons in conjunction with generative artificial intelligence models and prompt statements, replacing a large number of hard-coded templates and manual editing. Simultaneously, explicit matching of feature inputs ensures consistency between the generated results and the internal scoring logic, reducing semantic bias. The server reduces terminal-side implementation complexity through a centralized format checking module, allowing terminals to perform only lightweight input and display operations, improving the overall system's scalability and maintainability.

[0196] Furthermore, in an alternative implementation, the server can employ generative AI models with different architectures. For example, a pre-trained language model based on bidirectional encoding representation combined with a task-specific feedforward network can achieve stronger natural language understanding capabilities; or a lightweight distillation model can be used to reduce inference latency and memory consumption. The server can also maintain multiple dedicated fine-tuning models based on the language style and structural requirements of different funding aid schemes, and select the corresponding model based on the scheme identifier when invoked, thereby further improving the matching degree between the generated text and the target document format.

[0197] In another implementation, the server can employ unsupervised clustering algorithms or metric learning methods to learn features from a large number of historical successful application documents. These features are then input as additional conditions into a generative artificial intelligence model or scoring function. This allows the system to reference the structure and wording of existing successful samples when generating application documents and recommendation reasons for new users, thereby improving the overall document quality and the probability of acceptance. This algorithmic processing utilizing historical data features demonstrates the invention's improvement over data management and computation processes within the computer, rather than simply formalizing human experience directly.

[0198] Through the aforementioned modular construction and replaceable implementation forms, the system of this invention forms an integrated data flow within the server, encompassing natural language input, feature extraction, structured data construction, vectorized matching, document template filling, and generative AI-driven description generation. The server employs specific data structures and algorithms at each stage to reduce redundant computation, optimize storage access, and lower communication load, thereby improving processing speed, matching accuracy, and overall system performance. Therefore, this invention does not merely automate human work; rather, it improves the technical performance of the information processing system itself by synergistically integrating generative AI models with traditional database technologies through specific computer internal structures and algorithmic processes.

[0199] use Figure 12 The processing flow is explained.

[0200] Step 1: The user enters the activity details on the terminal and sends it. In the terminal's graphical interface, the user enters text about their activity in a text input box using a keyboard or touchscreen, such as "I want to implement a new cashless payment system in my local store to improve checkout efficiency." The terminal uses this text as input data and encapsulates it, along with the user's identifier and a timestamp, into a request message. The terminal performs data processing including: encoding the text (e.g., UTF-8), escaping special characters, and organizing the data into a key-value pair structure. The terminal then sends the processed request to the server via a communication network using the HTTP or HTTPS protocol. The input for this step is the raw natural language text entered by the user on the terminal, and the output is a network request message containing text information and the user's identifier.

[0201] Step 2: The server receives text messages and stores the raw data. The server receives request messages from the terminal at the network interface. In its communication management module, the server parses the messages, reading the request header and body, and extracting the activity content text and user identifier from the request body. The server uses this text information as input and performs data processing including: removing leading and trailing spaces, using Unicode encoding, checking for empty text, and recording the reception time. The server writes the processed text and the associated user identifier into the raw text information table in the database, generating a record containing a primary key ID. The input to this step is the text field and user identifier from the network request; the output is the raw text record stored in the database and its unique record ID.

[0202] Step 3: The server constructs and parses prompt statements and calls a generative artificial intelligence model for natural language parsing. The server reads the raw text record stored in step 2 from the database. Using this text as input, the server constructs a prompt statement by combining it with a predefined parsing task template. For example, the server generates the prompt statement: "Please extract the industry category, project type, company size information, and 3 to 5 keywords from the following project description: '[User Input Text]'." The server concatenates this prompt statement with the raw text to form the input sequence for the generative AI model. The server invokes the generative AI model deployed on a local or remote computing platform, encodes the input sequence into token IDs, performs embedding mapping and multi-layer self-attention operations, calculates the semantic vector for each token, and predicts the industry label, project type, size information, and keyword list through the model's output layer. The server post-processes the model output, selecting the labels and keywords with the highest probabilities to form the parsing result object. The input for this step is the raw text and the parsing task prompt statement; the output is a structured parsing result containing industry classification, activity type, size information, and a list of feature words.

[0203] Step 4: The server generates structured data objects based on the parsing results and stores them in the database. The server takes the parsed result object obtained in step 3 as input and constructs a unified data structure in the structured data generation module, such as records containing fields like "Industry Classification," "Activity Type," "Scale Parameter," "Keyword List," and "Regional Information." The server performs data processing including mapping text tags to standardized codes (e.g., industry codes), mapping scale descriptions to numerical ranges, and deduplicating and sorting the keyword list by importance. The server stores this structured data record in a structured data table in the database and establishes a foreign key relationship with the original text record ID. The input to this step is the parsed result object, and the output is the structured data record written to the database and its unique ID, used for subsequent matching of funding assistance programs.

[0204] Step 5: The server retrieves funding assistance information based on structured data and calculates evaluation values. The server reads the structured data records generated in step 4 from the structured data table. Using the industry classification, activity type, scale parameters, and keyword list as input, the server accesses the funding assistance policy table. The data operations performed by the server include: constructing SQL query conditions based on industry classification and activity type to filter basic candidate policies; encoding the structured data into query vectors and reading the pre-stored feature vectors of each policy from the funding assistance policy table; calculating the similarity (e.g., cosine similarity) between the query vectors and each policy vector, and performing range judgment on the scale parameters and the policy scale range. The server uses a preset weighting function to weight and sum the industry fit, activity fit, and condition matching to generate an evaluation value for each candidate funding assistance policy. The input for this step is the structured data records and the policy features from the funding assistance policy table; the output is a list of candidate funding assistance policies with evaluation values.

[0205] Step 6: The server selects the optimal financial aid system and generates a recommendation message. The server takes the candidate list and its evaluation values ​​obtained in step 5 as input. Based on the ranking results, the server selects one or more financial assistance programs with the highest evaluation values ​​as the optimal financial assistance program. The server records the matching details of each program in the matching module (matching industry fields, project type, whether the scale falls within the range, keyword overlap, etc.). Based on these matching details, the server constructs a recommendation reason prompt statement for use by the generative AI model, for example: "Please generate a recommendation reason explaining why this financial assistance program is suitable for users based on the following matching information: industry perfectly matched, project type is cashless payment system import, enterprise scale is SME, program supports SME digitalization projects." The server outputs this prompt statement to the generative AI model calling module. The input of this step is the candidate program list and matching details, and the output is the selected optimal financial assistance program and the prompt statement used to generate the recommendation reason.

[0206] Step 7: The server invokes a generative artificial intelligence model to generate a text explaining the reasons for the recommendation. The server takes the recommendation prompts generated in step 6 as input text and generates the text using a generative AI model in decoding mode. The server tokenizes and embeds the prompts, and generates recommendation sentences sequentially during the decoding phase, using temperature parameters and maximum length limits to control output diversity and length. The server collects the generated natural language recommendation text and removes leading and trailing whitespace to form the recommendation string. The input for this step is the recommendation prompt, and the output is the recommendation text explaining the suitability of the optimal financial assistance system, which is displayed to the user.

[0207] Step 8: The server selects a document template and fills in structured fields to generate the initial version of the application documents. The server takes the selected optimal funding assistance system identifier and the structured data records from step 4 as input, and loads the corresponding application document template from the document template table. The server parses the template in the document template management module, identifying placeholders for each predetermined item, such as "Project Name," "Project Summary," and "Budget Amount." Based on industry classifications, activity types, scale parameters in the structured data, and important sentences in the original text, the server performs field mapping and assignment for these placeholders; for example, it extracts project name phrases from the text and fills them into the "Project Name" field. The server marks the source information (obtained through parsing or extracted from the original text) for the automatically populated fields and generates an intermediate document object containing all field values. The inputs to this step are the funding assistance system identifier, structured data, and template definition; the output is the initial version of the application document data before the generation of the lengthy explanatory text.

[0208] Step 9: Server configuration description text prompts and generate purpose description paragraphs, etc. The server takes the incomplete fields (such as purpose description, background description, and effect description) from the initial version of the application document generated in step 8 as input. For each field, the server constructs a prompt statement for the generative AI model. For example, the server generates: "Based on the following project summary, please generate a project background description suitable for a funding application, approximately 400 words long, in a formal tone: '[Project Summary Field Content]'." The server inputs these prompt statements into the generative AI model, which performs autoregressive generation based on the prompt statements, outputting the corresponding explanatory text paragraphs. The server associates each generated explanatory text with the corresponding field and writes it into the intermediate document object. The input for this step is the incomplete field content and the prompt statements for generating the explanations; the output is the initial version of the application document data complete with the purpose description, background description, and effect description.

[0209] Step 10: The server sends the initial version of the application documents and reasons for recommendation to the terminal. In the user interface management module, the server takes the initial version of the application document generated in step 9 and the recommendation reason text generated in step 7 as input. The server organizes these two into a unified response structure, including field names, current values, editability flags, and the recommendation reason string. The server performs data serialization, converting the structure into a format that can be transmitted over the network (e.g., a field list in JSON format), and sends it to the terminal via HTTP or HTTPS response. The input for this step is the initial version of the application document and the recommendation reason text; the output is a network response message carrying this content, which is received and displayed by the terminal.

[0210] Step 11: The terminal displays the application documents and receives user correction input. After receiving the server's response, the terminal parses the field list and recommendation reason text locally, rendering each field as a visual form component, such as text boxes, dropdown lists, and labels. The terminal displays the recommendation reasons on the page for user reference. Users view the initial version of the application document on the terminal, entering or modifying fields, such as filling in budget amounts or correcting project names. The terminal performs basic format checks while the user edits, such as limiting character counts and prohibiting empty values. The terminal collects the user-corrected field values ​​into an updated data structure. The input to this step is the initial version of the application document data and recommendation reasons returned by the server; the output is a set of updated fields containing the user's corrections.

[0211] Step 12: The terminal sends the corrected content, the server performs verification, and updates the final version of the application document data. After the user confirms the editing, the terminal sends the updated field set generated in step 11 as input to the server via a network request. The server receives this request in its verification and error checking module, parses the updated fields, and merges them with the initial version data of the original application document. The server performs system-level verification on the merged application document data, including checking whether required fields are filled, verifying whether numerical fields are valid, and checking the date format and logical order. Based on the verification results, if an error is detected, the server generates an indication message containing the error item and explanation and sends it back to the terminal; if the verification passes, the merged application document data is marked as the final version and written to the application document table in the database. The input to this step is the set of corrected fields sent by the terminal and the initial version data of the original application document; the output is the updated final version application document data and (if errors are found) the corresponding error indication message.

[0212] Step 13: The server converts the final version of the application document data into a target electronic document or data structure and submits it to an external examination facility. In the external examination device interface module, the server takes the final version of the application document data generated in step 12 as input. The server selects the target format according to the requirements of the external examination device, such as generating a portable document file or a specific markup language structure. The server calls the document generation library or serialization module to render the field content into the formal document template and performs format processing such as layout and page numbering. Subsequently, the server compresses and encrypts the generated electronic document or data structure, and then sends it to the external examination device via a predefined communication protocol (such as HTTPS API calls or secure file transfer protocols). The server records the submission result and the returned acceptance number. The inputs to this step are the final version of the application document data and the interface specifications of the external examination device; the outputs are the successfully submitted electronic document or structured data, along with the corresponding submission status and acceptance information.

[0213] Alternatively, an emotion engine for inferring user emotions can be combined. That is, the specific processing unit 290 can also use the emotion-specific model 59 to infer user emotions and perform specific processing using user emotions.

[0214] Example 2 The flow of a specific process in Example 2 will be described. Each part of the system described below is implemented by the data processing device 12 and the smart device 14. The data processing device 12 will be referred to as the "server," and the smart device 14 as the "terminal."

[0215] Existing technologies for applying for financial aid typically rely on rule matching or simple keyword retrieval for policy selection and document generation. These technologies suffer from several problems: First, the server-side stores and retrieves user-inputted project descriptions using fixed fields, lacking a deep understanding of the semantics of natural language text. This results in matching aid policies based on manually set rules, leading to poor scalability, high maintenance costs, and decreased matching accuracy when policy terms are updated or project types diversify. Second, existing systems often use fixed templates to directly fill in a few fields when generating application documents. The server cannot dynamically adjust the chapter structure and text content according to different policy requirements, resulting in low-quality automatically generated documents that require extensive manual editing by users on their devices, reducing overall processing efficiency. Furthermore, even when generative AI models are introduced into traditional systems, they are often only invoked in a single stage, such as for writing explanatory text. The server's construction of prompts is one-off and unstructured, lacking a mechanism to decompose and drive the generative AI model in stages through multiple sub-tasks such as "domain determination," "system matching scoring," and "chapter generation." Therefore, the model's reasoning and generation capabilities cannot be fully utilized, and it is difficult to form a reusable and scalable prompt control process. In addition, existing technologies for integrating generated results with document formatting on the server side are rather crude. Servers often only provide simple text or general document formats, lacking a system for automatically synthesizing fixed-format electronic documents through markup language templates and enabling bidirectional interaction with structured text. When users edit on the terminal, the server struggles to update the formatted document and structured data in a timely and consistent manner, thus hindering improvements in the overall system's automation, maintainability, and user experience.

[0216] The specific processing performed by the specific processing unit 290 of the data processing apparatus 12 in Embodiment 2 is achieved by the following means.

[0217] In this invention, the server includes: a device for receiving project-related information from a user terminal and registering it in a storage device; a device for constructing multi-stage prompt statements for a generative artificial intelligence model based on the registered information and calling the generative artificial intelligence model to obtain parsing results containing domain information and multiple important words; a device for retrieving multiple funding assistance system candidates from the storage device based on the parsing results and calculating a suitability index to select a funding assistance system; a device for constructing prompt statements for each component of the application document based on the application conditions of the selected funding assistance system and the parsing results, and obtaining application document text data containing multiple chapters through a generative artificial intelligence model; a device for automatically embedding the obtained text data into a markup language format document generation template to generate a fixed-format electronic document; a device for providing structured text data corresponding to the electronic document to the user terminal through a user interface and receiving user-edited content to synchronously update the structured text data and the electronic document; and a device for storing the updated electronic document as the final version and outputting application processing completion information. This allows for the formation of a server-side pipeline for processing financial aid applications. This pipeline is centered on a generative artificial intelligence model and is driven by sophisticated multi-stage prompts. It enables high-precision semantic parsing of natural language project descriptions, intelligent matching and scoring of financial aid programs, automatic generation of multi-chapter application documents, and the generation of fixed-format electronic documents through the linkage of markup language templates and structured data. This significantly improves the system's document generation quality, processing efficiency, and consistency with user terminals, thereby achieving an overall performance improvement and resource utilization optimization for computers in financial aid application processing.

[0218] "System" refers to a computer-based technical solution consisting of at least one server, at least one user terminal, and a storage device for storing data, interconnected through a communication network, to perform the processing of financial assistance applications.

[0219] A "server" is a computing device equipped with a processor and memory, capable of running programs to receive, process, and output data related to financial assistance applications, and interacts with user terminals and storage devices through a communication interface.

[0220] "User terminal" refers to a computing device operated by a user for inputting project-related information and viewing and editing application documents, including but not limited to computers, tablets, or mobile terminals.

[0221] "Storage device" refers to data storage resources used to store project information, funding assistance system information, analysis results, application document data, and electronic documents in a searchable manner, including but not limited to database systems, file storage systems, or other non-volatile storage media.

[0222] "Project-related information" refers to descriptive data entered by users regarding the proposed activities or plans, including but not limited to text or structured data describing the characteristics of the project such as project name, project purpose, activity content, schedule, implementation location, and the number of participants.

[0223] "Generative AI model" refers to an AI model trained based on machine learning or deep learning techniques that can generate output text or structured data from input text. It is used to perform semantic parsing of project descriptions, extract information, and generate document content.

[0224] "Prompt statements" refer to text content that is constructed by the server and sent to the generative artificial intelligence model to instruct the model to perform specific parsing or generation tasks, including descriptions of the input data and limitations on the output format and task requirements.

[0225] "Analysis results" refer to the analytical data output by the generative artificial intelligence model based on project-related information and prompts, including at least domain information, multiple key words, and optional summary information or other structured features.

[0226] "Domain information" refers to the classification information extracted from the project description to indicate the technical or business category to which the project belongs, such as environmental protection, education, healthcare, or regional development.

[0227] "Important words" refer to key terms extracted from the project description that can characterize the core content of the project and are used as a matching basis when searching for financial assistance systems and calculating suitability indicators.

[0228] "Candidates for financial assistance programs" refers to a portion of multiple program records registered in the storage device that can provide financial support for projects. Each record contains attributes such as applicable areas, application conditions, and budget limits, which are used to match projects.

[0229] The "suitability index" is an evaluation value calculated by the server based on the correspondence between the project analysis results and the candidates for each funding assistance system. It is used to quantitatively represent the degree of matching between the project and each candidate system.

[0230] "Application requirements" refer to the various restrictions and requirements stipulated by the funding assistance system that a project must meet when applying for the system, including but not limited to the applicant's qualifications, project scope requirements, budget restrictions, and time requirements.

[0231] "Application document text data" refers to the text content of multiple chapters used to constitute a funding assistance application, including sections such as project overview, implementation plan, budget details, and expected results.

[0232] "Constituent elements" refer to the content parts in the application documents that are regarded as independent generated or edited units, including chapters, paragraphs or fields, which are used to decompose the generation task of the generative artificial intelligence model.

[0233] "Document generation description format" refers to a markup language or other textual document template language used to describe the structure and layout of a document. By embedding text data in it, electronic documents with a predetermined layout can be generated.

[0234] "Markup Language Format" refers to a text format that uses markup symbols to describe document structure, paragraph hierarchy, and layout attributes, and is used as a template for generating electronic documents on a server.

[0235] "Fixed-format electronic document" refers to an electronic file format that is basically fixed in layout when presented on different terminals after being processed by the document generation description format. It is used as a formal application document for submission or archiving.

[0236] "Structured text data" refers to application document content data managed by the server in a field or hierarchical manner. It organizes chapters, paragraphs or project elements in an editable and searchable form, which facilitates synchronous updates and document generation.

[0237] The "user interface" refers to the interactive interface that provides users with access to information display and editing, including web pages or application interfaces running on user terminals, used to display structured text data and electronic documents, and to receive user editing operations.

[0238] "Application processing completion information" refers to the status information used to indicate that the application process for funding assistance for a specific project has been completed in the system, including but not limited to data such as submission completion notification, acceptance confirmation number, or processing result status.

[0239] In this embodiment of the invention, a server, a terminal, and a user collaborate to implement a financial assistance application processing system. The server is typically deployed on a computer device with a multi-core processor and large-capacity main memory, and the operating system can be a Unix-like system. The server runs relational database management software, such as a relational database management system, on the storage device, and runs a scripting language runtime environment, such as an interpreted language environment, and a document typesetting system, such as a markup language-based typesetting toolkit. The server establishes a communication channel with the user terminal based on the HTTPS protocol through a communication interface. The terminal can be a personal computing device or a mobile computing device with a browser application installed.

[0240] The server maintains multiple data table structures in the storage device. The server stores project information in the project data table, which includes project identifier fields, text description fields, and structured attribute fields (such as date, location, and number of participants). The server stores funding assistance policy information in the policy data table, which includes policy identifier fields, applicable fields, keyword fields, application condition fields, and application template definition fields. The server stores parsing results and application document drafts in the parsing results table and document draft table, respectively. The data tables are linked through foreign key fields, thus achieving structured links between projects, parsing results, policy candidates, and application documents.

[0241] The server is divided into several sub-modules in the application layer, including a parsing module, a matching module, a document generation module, a document formatting module, and a user interface service module. The parsing module is responsible for interacting with the generative artificial intelligence model, the matching module is responsible for calculating the suitability index between the project and each policy candidate, the document generation module is responsible for generating application document text by chapter, the document formatting module is responsible for generating fixed-format electronic documents using markup language templates, and the user interface service module is responsible for interacting with the terminal for structured text data.

[0242] The server invokes a generative artificial intelligence model in its parsing module. This model is preferably based on a neural network architecture with a multi-layered self-attention structure, such as a sequence-to-sequence model with multiple encoders and decoders. Model parameters include multi-head attention weight matrices, feedforward network weight matrices, and embedding vector matrices. During inference, the server concatenates item-related information with pre-constructed prompts into an input sequence, which is then segmented and vectorized before being fed into the generative AI model. Internally, the model calculates the correlation between input tokens using a self-attention mechanism and extracts high-dimensional features through multi-layer stacking. Finally, it generates an output sequence containing domain information, key words, and a summary text. The server then parses this output sequence to extract structured domain labels and keyword lists.

[0243] During the training phase, the server can fine-tune the generative AI model using historical funding application datasets. The server uses historical project descriptions, approved funding scheme labels, and standardized application documents as training samples. Through supervised learning, the server sets a cross-entropy loss function or a sequence-to-sequence loss function to measure the differences between the model's output domain labels, keyword sets, and document text and the labeled results. The server updates the model's weight parameters using backpropagation, performing weight updates in each training epoch using batch gradient descent or adaptive learning rate optimization algorithms. To improve the model's generalization ability, the server can introduce data augmentation during the training phase, such as synonym replacement and sentence perturbation, to generate diverse training samples, thereby reducing reliance on specific expression methods.

[0244] The server calculates a suitability index for candidate financial aid programs in its matching module. The server reads multiple key words from the parsed results and compares them with the keyword field of each financial aid program record in the program data table. The server uses a vector space model to map keywords to a vector space and obtains a preliminary matching score by calculating cosine similarity or weighted similarity. Simultaneously, the server can construct specific prompts, taking project descriptions and candidate program summaries as input, and instructing a generative AI model to output a matching score and rationale text for each candidate program. The server internally sets weighting rules to synthesize the keyword-based similarity and model-based inference scores according to preset weights, thereby obtaining the final suitability index. Compared to single-keyword rules, this combined algorithm can significantly reduce matching errors caused by differences in terminology on the server side, improving the accuracy of financial aid program selection.

[0245] When constructing the prompt, the server employs a phased, multi-tasking, unconventional control approach, rather than issuing a highly abstract request all at once. The server first constructs a prompt for "project resolution," for example: "You are a funding system matching assistant. Please read the following project description and output: 1) Project area (e.g., environmental protection, education, medical care, regional development, etc.); 2) 3-10 keywords that can represent the core content of the project; 3) A one-sentence summary of the project purpose. The project description is as follows:..." The server then constructs a prompt statement for "system scoring" during the matching phase, for example: "Below is a project description and brief information on several funding assistance programs. Please rate each program on a scale of 0 to 100 based on how well the project matches the program, and provide a one-sentence explanation. Project Description: ... List of Funding Assistance Programs: ...." During the document generation stage, the server then constructs prompts for "chapter generation," such as: "Based on the following project description, write the 'Project Overview' section of the government funding application, approximately 500 words in length. The tone should be formal, highlighting the project's background, purpose, and necessity. Project description: ……." And prompts used in the "Implementation Plan" section, such as: "Based on the following project description, generate a phased implementation plan, including timelines and key tasks, approximately 400 words. Project description: ……." The aforementioned phased prompt design enables the server to call the generative artificial intelligence model separately for different subtasks. The output of each phase is explicitly used by the subsequent phases, forming a clear data flow, thereby realizing a traceable and tunable model call process on the server side.

[0246] In the document generation module, the server writes the text of each chapter obtained from the generative AI model into a document draft table. The server treats each chapter as a separate component and assigns it a chapter identifier to support subsequent individual editing and regeneration of each chapter. When the user makes partial modifications, the server can only re-invoke the generative AI model to polish the modified chapters, without having to regenerate the entire application document, thereby reducing computational load and improving the processing speed of document updates.

[0247] The server uses a markup language format as the description format for document generation in its document formatting module. The server pre-installs multiple document template files in its storage device. These templates define layout attributes such as chapter order, heading styles, paragraph indentation, headers, and footers using markup symbols. The server maps structured text data to placeholder positions in the templates by chapter and calls the formatting tool to perform a compilation operation, generating a fixed-format electronic document file. Because the server separates structured content from layout description, it can quickly adapt to different submission format requirements by modifying templates without changing business logic, ensuring consistent layout across various terminals, thereby improving system maintainability and portability.

[0248] The terminal interacts with the server's user interface service module through a browser or local application. The terminal receives structured text data from the server and presents the content of each chapter as editable text areas on the interface. The terminal can also view preview images of fixed-format electronic documents or display the final layout through an embedded document viewer. When the user modifies the text on the terminal, the terminal sends the modified structured text data back to the server in a lightweight data format. This design reduces communication overhead because the terminal does not need to upload the entire electronic document file each time, but only transmits the modified text fields.

[0249] Users control the system through a user interface on the terminal. Users input project-related information, such as "This project aims to restore the ecological function of degraded forest land in the region through a three-year cycle of tree planting and forest management activities, planned to begin in a certain month of a certain year, with an estimated number of participants," etc. Based on this input, the server performs parsing, matching, document generation, and formatting according to the aforementioned data flow. Users can view the generated application draft and fixed-format electronic document multiple times on the terminal, gradually adding details such as "participant composition" and "safety management measures." The server performs incremental updates after each received modification data and retains multiple versions of the application document for retrospective review and comparison.

[0250] Through its modular structure and prompt-based control mechanism, the server not only automates the traditional manual application writing process but also establishes a new internal processing structure at the computer technology level. By controlling the generative AI model with multi-stage prompts, the server decomposes the complex natural language understanding and generation tasks into manageable sub-tasks, avoiding the unstable output caused by large-scale generation at once. This technically improves the quality of the output text and the matching accuracy. Furthermore, by combining the model's inference results with vector similarity calculations to form a composite fitness index, the server makes the matching algorithm robust to differences in textual expression, significantly outperforming traditional techniques based solely on rule-based or keyword-based Boolean retrieval.

[0251] Furthermore, the server employs a multi-table relational structure among projects, parsing results, policy candidates, draft documents, and electronic documents. This allows the system to quickly locate all intermediate data related to any project, eliminating the need for full table scans during partial updates or repetitive generation, reducing database access and improving overall processing speed. In terms of terminal interaction, the server uses a dual-track structure that maintains structured text and fixed-format electronic documents in parallel. This ensures that user editing operations are only performed on structured text, while formatting is triggered only when necessary, thereby improving computational efficiency and reducing system load.

[0252] In one alternative implementation, the server can use different types of generative AI models, such as encoder-based text classification models for domain and keyword extraction, and decoder-based language models for chapter text generation. In another implementation, the server can dynamically adjust the granularity of prompts based on the complexity of the funding assistance program; for example, in highly segmented domain scenarios, it can add the extraction of features such as "geographical constraints" and "target group characteristics." In yet another implementation, the server can introduce a rule engine module to cross-validate manually defined hard constraint rules with the model output, filtering candidate programs that clearly do not meet the specified conditions, thus combining rule logic with statistical model reasoning.

[0253] Therefore, through the comprehensive configuration of specific software module division, data structure design, generative artificial intelligence model control strategy, and document typesetting and terminal interaction mechanism, the system of the present invention achieves higher semantic understanding accuracy, faster matching and document generation speed, lower communication and computing load, and stronger scalability when processing financial assistance application business, thus achieving substantial technical effects in the field of computer technology.

[0254] use Figure 13 The processing flow is explained.

[0255] Step 1: The user enters project information on the terminal and sends it. Users open the page provided by the server on their terminal through a browser or local application, and fill in text fields such as project name, project purpose, activity content, start time, implementation location, and number of participants.

[0256] Input: The various fields entered by the user in the terminal interface.

[0257] After the user clicks the "Next" or "Submit" button, the terminal packages the form content into request data (such as JSON or form encoding) and sends an HTTP request to the server via the HTTPS protocol.

[0258] Output: Network request data containing the item field sent by the terminal to the server.

[0259] Step 2: The server receives project information and writes it to the storage device. The server receives an HTTP request from the terminal. The server's application parses the request body and converts each field into an internal data structure (such as key-value pairs).

[0260] Input: Network request data containing the project field.

[0261] The server performs basic preprocessing on the text (removing spaces, standardizing date formats, and checking required fields). Then, the server calls the database interface to insert a new record in the project data table, writing the project name, project purpose, activity content, etc. into the corresponding fields, and generating a project identifier.

[0262] The server stores the item identifier in memory as an index for subsequent processing.

[0263] Output: A new item record in the storage device and its corresponding item identifier.

[0264] Step 3: The server constructs a project parser that uses prompts and calls a generative artificial intelligence model. The server reads the project record that was just stored from the project data table and combines the text fields such as project name, purpose, and activity content into a project description.

[0265] Input: Text field data from the project record.

[0266] The server constructs a project parsing prompt statement based on a preset template, for example: "You are a funding system matching assistant. Please read the following project description and output: 1) Project area (e.g., environmental protection, education, medical care, regional development, etc.); 2) 3-10 keywords that can represent the core content of the project; 3) A one-sentence summary of the project purpose. The project description is as follows:..." The server concatenates the prompts and project descriptions into the model input text, which is then sent to the generative AI model via the model interface. Internally, the model performs word segmentation, embedding, and multi-layer attention calculations on the text, outputting domain labels, a keyword list, and a summary of the project's objectives.

[0267] Output: The parsed text received by the server from the model interface, containing domain information, several key terms, and a summary of the project objectives.

[0268] Step 4: The server parses the model output and generates structured analysis results. The server parses the parsing results text returned by the generative artificial intelligence model, identifies parts such as "domain information", "keyword list" and "project purpose summary", and breaks these contents down into structured data.

[0269] Input: The parsed text containing domain information and key words.

[0270] The server writes the domain label into the "domain field" of the parsing results table, splits the keyword list into multiple entries and writes them into the "keyword field" or the association table, and writes the project purpose summary into the "summary field". At the same time, it saves the corresponding project identifier in the parsing results record.

[0271] Output: A structured parsing result record for this project in the storage device.

[0272] Step 5: The server retrieves candidates for funding assistance programs based on the parsing results. The server reads the domain information and keyword list of the project from the parsing result table.

[0273] Input: Domain information and several key words from the parsed results.

[0274] The server performs a retrieval operation in the policy data table: on the one hand, it filters out policy records that match the domain based on the domain field; on the other hand, it performs text matching or vector similarity calculation in the keyword field to obtain several candidates for financial assistance policies.

[0275] The server temporarily stores the candidate system's identifier, name, summary, and other information into an in-memory data structure or candidate table and associates it with the project identifier.

[0276] Output: A set of candidate financial assistance programs associated with this project and their basic attributes.

[0277] Step 6: Suitability Indicators for Server Computing Funding Assistance Program Candidates The server reads the candidates for financial assistance programs and their attributes obtained in step 5, and also reads important words from the project analysis results.

[0278] Input: A candidate set of financial assistance programs and their attributes, and a set of key terms for each program.

[0279] The server first calculates the preliminary similarity between each candidate system and the project based on keyword vectors or word frequency weights, for example, by calculating keyword overlap or cosine similarity to obtain a numerical score.

[0280] Then, the server constructs a second type of prompt statement, for example: "Below is a project description and brief information on several funding assistance programs. Please rate each program on a scale of 0 to 100 based on how well the project matches the program, and provide a one-sentence explanation. Project Description: ... List of Funding Assistance Programs: ...." The server combines the project description and a summary of each candidate policy into the prompt statement and sends it to the generative AI model. The model then outputs a matching score and a rationale text for each candidate policy.

[0281] The server combines the keyword similarity score and the matching score given by the model according to a preset weight to synthesize the final fitness index, and writes it into the record corresponding to each candidate system.

[0282] Output: Suitability index and matching rationale for each candidate financial aid scheme.

[0283] Step 7: Server selection target funding assistance system The server sorts all candidate systems from highest to lowest score based on the suitability index calculated in step 6.

[0284] Input: A set of candidate financial aid schemes with a fitness index.

[0285] The server can select the highest-scoring system, or several systems with scores exceeding a threshold, as the target funding assistance system. The server writes the identifiers of these selected systems into the associated record corresponding to the project identifier.

[0286] Output: One or more selected funding assistance regime identifiers associated with this project.

[0287] Step 8: The server constructs prompts for generating application documents and calls a generative artificial intelligence model to generate the text for each chapter. The server reads the application requirements and document structure definition of the selected funding assistance program from the program data table (e.g., it must include sections such as "Project Overview", "Implementation Plan", "Budget Details", and "Expected Results"), and reads the project description and analysis results from the project data table and the analysis results table.

[0288] Input: Application requirements and document structure for the target funding assistance program, project description, and analysis results.

[0289] The server constructs corresponding prompts for each chapter, for example: "Based on the following project description, write the 'Project Overview' section of the government funding application, approximately 500 words in length. The tone should be formal, highlighting the project's background, purpose, and necessity. Project Description: ... Application Requirements: ...." "Based on the following project description and application requirements, generate an 'Implementation Plan' section, including time phases and main tasks, approximately 400 words in length. Project Description: ... Application Requirements: ...." The server sends these prompts sequentially to the generative AI model, which returns the complete text of the corresponding chapter. The server then stores the text of each chapter along with its chapter identifier in a document draft table.

[0290] Output: A collection of application document text data stored in the storage device, organized by chapter.

[0291] Step 9: The server generates a document template, populates it with data, and uses a typesetting tool to generate a fixed-format electronic document. The server reads all chapter text from the document draft table and maps it to placeholders in the document template based on the chapter identifier.

[0292] Input: Application document text data stored by chapter.

[0293] The server inserts the chapter text into a markup language template file to form a complete document description. The server then calls a document formatting tool to perform a compilation operation, converting the markup language description into a fixed-format electronic document file.

[0294] The server registers the path of the generated electronic document file in the document record and retains the corresponding version number.

[0295] Output: Fixed-format electronic document files associated with the project and updated document records in the storage device.

[0296] Step 10: The server sends structured text data and document preview information to the terminal. The server reads the structured text of each chapter from the document draft table and obtains the access path of the corresponding fixed-format electronic document.

[0297] Input: Structured chapter text data and electronic document paths.

[0298] The server generates response data for terminal display through the user interface service module, organizes the text of each chapter into an editable field structure, and sends it to the terminal along with an electronic document download or preview URL.

[0299] Output: Response data containing editable chapter content and electronic document preview information.

[0300] Step 11: The terminal displays the application documents and allows users to make modifications. After receiving the response data from the server, the terminal displays the application document content chapter by chapter on the interface, placing the text of each chapter into an editable area; the terminal also provides an entry point for viewing or downloading the electronic document on the interface.

[0301] Input: Chapter text and electronic document preview information received from the server.

[0302] Users can read the contents of each chapter on the terminal interface. If they find any parts that need to be modified, they can directly edit them in the corresponding text area, adding or correcting the text, such as adding budget details or modifying the time plan.

[0303] When the user clicks the "Save Changes" or "Update Draft" button, the terminal packages the edited text of each chapter into updated data and sends it to the server via HTTPS.

[0304] Output: Contains update request data for the chapter text modified by the user.

[0305] Step 12: The server updates the structured text and selectively invokes a generative artificial intelligence model for polishing. The server receives update request data sent by the terminal and parses out the modified chapters and their new text content.

[0306] Input: The chapter text data modified by the user.

[0307] The server writes the new text into the corresponding section field of the document draft table, replacing the old version content; the server determines whether automatic polishing is needed based on the settings, and if so, the server constructs polishing prompts for the modified section, such as: "Please refine the following text without altering the factual content, making the language more rigorous and in line with the style of a government funding application: Original text: ..." The server sends the prompt and the original text to the generative artificial intelligence model, receives the polished text, performs a simple semantic and format check, and then writes it back to the document draft form.

[0308] Output: Updated structured chapter text data and (if applicable) polished chapter text.

[0309] Step 13: The server regenerates the fixed-format electronic document and stores the final version. After the document draft table is updated, the server reads the latest text of all chapters from it again.

[0310] Input: The updated structured text data for all chapters.

[0311] The server re-populates the text into the markup language template and calls the typesetting tool again to generate a new, fixed-format electronic document file. The server writes the new file path and version number into the document record and marks the version as "final" or "submitted".

[0312] Output: The updated final version of the fixed-format electronic document and the corresponding version history.

[0313] Step 14: The user confirms the final document on the terminal and issues a submission command. Users can view the latest chapter text and electronic document preview returned by the server on the terminal interface. After confirming that the content is correct, users can click "Submit Application" or a similar button on the interface.

[0314] Input: The final version of the document displayed on the terminal and the user's operation instructions.

[0315] The terminal packages the project identifier and the "submit" operation instruction into a submit request data and sends it to the server via HTTPS.

[0316] Output: Submission command data sent from the terminal to the server.

[0317] Step 15: The server registers the application submission status and outputs an application processing completion message. The server receives the submission command data sent by the terminal and parses out the project identifier.

[0318] Input: Submission command data containing the project identifier.

[0319] The server updates the status field of the corresponding project in the project data table to "Submitted" or a similar status, and records the submission time and the final electronic document version number. The server generates application processing completion information, such as "Application successfully submitted, number X, estimated review period is Y days".

[0320] The server returns the completion information to the terminal through the user interface service module; the server can also send notifications to the user's pre-registered contact information through the message sending module.

[0321] Output: The project submission status written to the storage device and the application processing completion information returned to the terminal.

[0322] Application Example 2 The process flow corresponding to the specific processing in Use Case 2 will be described below. The various parts of the system described below are implemented by the data processing device 12 and the intelligent device 14. In addition, the data processing device 12 is referred to as the "server" and the intelligent device 14 is referred to as the "terminal".

[0323] Existing technologies supporting financial aid or grant applications typically employ rule matching or fixed templates, selecting a program from user-inputted project descriptions and generating application documents. This approach suffers from the following technical shortcomings.

[0324] First, when processing unstructured natural language text, servers mostly perform simple keyword retrieval or fixed field extraction, and cannot obtain multi-dimensional structured information such as project type, purpose, scale, budget, period, and target field through deep natural language parsing. This results in low accuracy of subsequent database retrieval and matching operations, and the server-side policy selection logic lacks intelligence and scalability.

[0325] Second, when interacting with generative AI models, servers often simply input the raw text directly into the model without constructing task-oriented structured prompts or introducing a combination of structured project information and institutional requirements. This makes it difficult for generative AI models to converge to an output that precisely corresponds to the target system in a timely manner, resulting in redundant, missing, or inconsistent generated content with the institutional conditions, which reduces the overall processing efficiency and stability of the server.

[0326] Third, during the process of generating and adjusting application documents, the server typically does not analyze the user's emotional state, nor does it feed back the emotional analysis results into the construction of prompts and the document generation process. It cannot adaptively adjust the tone, level of detail, and emphasis of the document according to the user's state of tension, anxiety, pressure, or reassurance, resulting in a poor user experience and requiring repeated manual modifications, which increases the number of data round trips and resource consumption between the server and the terminal.

[0327] Fourth, the server lacks an integrated process control mechanism and cannot complete the entire process of user information reception and storage, natural language parsing and structured processing, support system candidate retrieval and scoring selection, automatic construction of prompts and emotional constraints for generative artificial intelligence models, format conversion after document generation, and external electronic submission under the same technical framework. Therefore, it cannot fully utilize the advantages of server processing and communication resources.

[0328] Therefore, it is necessary to provide a new computer implementation that enables servers to perform text parsing, policy matching, prompt generation, emotional feedback control, and document generation and submission in a more efficient and intelligent manner, thereby substantially improving the processing power and operational efficiency of computer technology itself in this type of business.

[0329] The specific processing performed by the specific processing unit 290 of the data processing apparatus 12 in Application Example 2 is achieved by the following means.

[0330] In this invention, the server includes a module for receiving information related to a matter and information related to an emotional state from a user and storing the information in a storage device; a module for performing natural language processing on the information related to the matter and extracting structured information including the type, purpose, scale, budget, period, and target area of ​​the matter, as well as multiple key pieces of information; a module for retrieving a system information storage device storing multiple support systems based on the structured information and the multiple key pieces of information and calculating the suitability score of each support system candidate to select the optimal support system; and a module for automatically generating a system information storage device that is input into the generator based on the structured information and the system requirements information of the selected support system. The system includes a module for generating application documents, a module for adding additional prompts to the prompts based on the user's emotional state and the results of emotional analysis, to automatically adjust the content or wording of the application documents, and a module for recording the application documents output by the generative AI model, converting them into electronic documents for external submission, and sending them to external review devices or support system providers via communication devices. This creates a technical chain within the server driven by structured information, using task-oriented prompts to control the behavior of the generative AI model, and providing closed-loop feedback of emotional analysis results to the document generation process. This improves the accuracy of natural language parsing and system matching, reduces the number of interactions and data redundancy between the server and the terminal, reduces the unnecessary computational overhead of calling the generative AI model, and automates and automates the entire application document generation and submission process within the computer system, thereby improving the resource utilization efficiency and overall performance of the computer in this type of business processing.

[0331] "User" refers to the entity that provides information about matters and emotional states to the system through a terminal, and confirms and corrects the generated application documents. It can be an individual or a representative of an organization.

[0332] "Event-related information" refers to natural language information entered by users for a project, activity, or plan, including unstructured or semi-structured data such as project name, background, purpose, scale, budget, period, implementation content, and target area.

[0333] "Emotional state-related information" refers to data used to represent a user's current psychological or emotional state, including the types of emotions that the user directly selects or inputs, as well as emotional characteristics implied in voice data, image data, or text data.

[0334] "Storage device" refers to a data storage resource used to electronically store user input information, parsing results, system information, prompts, and generated documents, including local storage media and remote data storage systems.

[0335] Natural Language Processing (NLP) refers to the technical process by which servers perform data processing operations such as text preprocessing, word segmentation, part-of-speech tagging, named entity recognition, syntactic analysis, and semantic analysis on information related to a matter, in order to extract structured and key information from natural language text.

[0336] "Structured information" refers to project attribute information represented in the form of fields, obtained through natural language processing and parsing, including data that can be directly used for calculation and retrieval, such as the type of matter, purpose, scale, budget, period, and target area.

[0337] "Key information" refers to keywords or key phrases extracted from information related to a matter that have a significant impact on the selection of support systems or the generation of documents, including elements such as domain terms, technical terms, amounts, locations, and times.

[0338] The “support system” refers to a set of rules provided by external organizations to provide funding, materials or other forms of support for projects or activities, including applicable conditions, support content, application requirements and review process.

[0339] "Institutional information storage device" refers to storage resources used to store structured data of multiple support systems, including information such as the applicable objects, upper limits of amount, fields, regions, application conditions, and time limits of each support system.

[0340] "Support system candidates" refers to a set of support systems that may be applicable to the matter, initially selected from the search results of the system information storage device based on the structured and key information of the project.

[0341] "Fitness score" refers to the numerical value calculated by the server based on the degree of matching between the structured information of the project and the requirements of the support system, using a predetermined algorithm or model. It is used to quantify the suitability of each support system candidate for the current matter.

[0342] The "optimal support system" refers to the target support system that is determined to have the highest overall match with the current issue after assessing and ranking multiple support system candidates based on their suitability.

[0343] "Institutional Requirements Information" refers to the set of necessary conditions and format requirements for an application corresponding to a certain support system, including required fields, required supporting documents, text structure, and length limits.

[0344] "Generative artificial intelligence models" refer to artificial intelligence models built on machine learning and deep learning technologies that can automatically generate coherent natural language text based on input text prompts.

[0345] "Prompt statements" refer to text instructions constructed by the server and input into the generative artificial intelligence model to instruct the model to perform specific generative tasks. These instructions include structured information about the project, supporting institutional requirements, and document generation requirements.

[0346] "Additional prompts" refer to supplementary text instructions that, in addition to basic prompts, incorporate conditions related to the user's emotional state, such as tone control, level of detail control, and emphasis configuration, to guide generative artificial intelligence models in adjusting the content or expression of documents.

[0347] "Sentiment analysis processing" refers to the computational process by which a server applies sentiment classification models or sentiment analysis services to information related to emotional states in order to identify and quantify the types and intensity of user emotions.

[0348] "Emotional category" refers to the category of user emotions identified through emotion analysis, including but not limited to emotional tags such as anxiety, stress, peace of mind, joy, and tension.

[0349] "Emotional intensity" refers to a numerical value or level that quantifies the type of emotion, reflecting the relative strength of a user's emotional state.

[0350] "Document tone" refers to the stylistic characteristics of the language used in application documents, including formality, reassuring tone, encouraging tone, directness, or euphemism.

[0351] "Detail level of description" refers to the depth and information density of the description of the project background, implementation content, risk countermeasures, etc. in the application documents, including different levels such as brief description and detailed description.

[0352] “Emphasis on the project” refers to the content elements that are highlighted or emphasized in the application documents, including the project’s advantages, social benefits, environmental benefits, and risk control measures.

[0353] "Application documents" refer to electronic texts that contain project descriptions, budget plans, expected outcomes, and other necessary information generated for the application of a particular support system, and are submitted to external review agencies.

[0354] "Electronic document format" refers to the file format that can be processed and transmitted by a computer system when application documents are submitted to external parties, including but not limited to text format, layout file format, or other structured electronic file format.

[0355] "Communication device" refers to the hardware and software components used to send and receive electronic data between a server and external devices, including network interfaces, communication protocol stacks, and related control programs.

[0356] "External review device" refers to a computing device or information processing system used to receive application documents and review applications for support systems.

[0357] "Support system providing device" refers to a computing device or information processing system used by the provider of the support system to manage support system information, receive applications, and provide feedback on results.

[0358] "User terminal" refers to an electronic device that allows users to interact with a server, including a display device, an input device, and communication functions, used to display application documents and receive user editing and instructions.

[0359] "Second application document" refers to the version of the application document that is regenerated by the server based on the user's modifications after the user modifies the application document initially generated by the generative artificial intelligence model, and is used for the final submission.

[0360] In one embodiment of the present invention, the server is installed in a computer rack inside a data center or factory, and the terminal is an electronic device (e.g., a mobile terminal, tablet terminal, or desktop terminal) with a display device and an input device, through which the user interacts with the server.

[0361] Servers can include multi-core central processing units, graphics processing units, large-capacity random access memory, and non-volatile storage media in terms of hardware. On the software side, servers run operating systems, database management systems, and various application components. The natural language processing and machine learning components of a server can be specifically composed of the following software: natural language processing libraries, such as a word segmentation and named entity recognition library, and a syntactic analysis library; machine learning frameworks, such as a tensor computation framework (corresponding to deep learning training and inference); generative artificial intelligence models, such as a text generation model based on a self-attention network structure; and sentiment analysis services, such as a sentiment analysis interface service.

[0362] In this invention, the server executes pre-stored programs to achieve collaboration among multiple functional modules. The server can be physically deployed in a modular fashion, and logically includes at least the following functional modules: a user information receiving and storage module, a text preprocessing and natural language parsing module, a feature extraction and structured information generation module, a support system retrieval and scoring module, a generative artificial intelligence model interface module, a prompt statement construction module, a sentiment analysis and sentiment feedback control module, a document generation and format conversion module, an external communication module, and a log and model management module. Data exchange between these modules occurs via an internal bus or message queue.

[0363] In the user information receiving and storage module, the server treats user input from the terminal as data objects with a unified structure. The server can use a relational database to store each piece of information as a record, with multiple fields such as raw text, sentiment input, parsing result, selected policy identifier, and generated document version. In this module, the server generates a unique identifier for each received request to ensure consistency in subsequent data processing and tracking.

[0364] In the text preprocessing and natural language parsing modules, the server performs specific data processing on the relevant information. The server can use natural language processing libraries to convert user-submitted natural language text sequences into labeled sequences. The server performs preprocessing operations such as sentence segmentation, noise removal, and standardization of numerical units; then it performs word segmentation and part-of-speech tagging to obtain word sequences and corresponding part-of-speech tags; next, it performs named entity recognition, extracting amounts, times, locations, organization names, and technical terms as entities; the server can also perform dependency parsing, representing sentences as dependency tree structures. Through vector operations and matrix multiplication within the model, the server embeds words into a high-dimensional feature space, then propagates them through multi-layer convolutional networks or multi-layer self-attention networks to generate context-sensitive semantic representations. In this process, the server does not simply match strings, but rather performs continuous-space calculations using numerical vectors and weight matrices to obtain high-dimensional features.

[0365] In the feature extraction and structured information generation module, the server transforms the features obtained from natural language parsing into structured information. The server can pre-train a multi-layer fully connected network or transformer encoder to classify sentence vectors or entity vectors, outputting labels such as item type, target domain, and priority. Simultaneously, based on rule templates, the server parses budget, period, and scale fields from the entity set; for example, it uses a unit normalization function to convert "500kW" and "30 million" into values ​​with uniform units. The server stores these structured fields in key-value pairs in the database, providing input for subsequent retrieval and scoring modules. Through this structuring process, the server significantly reduces fuzzy matching operations in subsequent retrieval calculations, improving index hit rate and computational efficiency.

[0366] In the support system retrieval and scoring module, the server utilizes structured information to perform fine-grained matching of various pre-stored support systems. The server can maintain a set of feature vectors for each support system, including vectors for applicable domains, budget ranges, regional codes, and object type codes. The server first uses conditional filtering queries to select an initial candidate set from the system information storage device; then, it inputs the project feature vectors and system feature vectors into a scoring network, which can be a multilayer perceptron or a matching network with attention weights, outputting a suitability score for each candidate system. The server sorts the scores and selects the one or more systems with the highest scores as the optimal support system. Because it employs vectorized matching and neural network scoring, the server can handle high-dimensional feature relationships and automatically learn the weights between different features, rather than relying solely on manual rules, thus maintaining matching accuracy and computational speed even under multiple constraints.

[0367] In the generative AI model interface module and prompt statement construction module, the server employs an explicit prompt statement construction algorithm to control the behavior of the generative AI model. The server first combines structured information with institutional requirements related to the selected system into an intermediate data structure, such as an ordered sequence containing field names and values. Based on a predefined template, the server converts these fields into natural language descriptions and inserts task instructions, output format requirements, etc., to form basic prompt statements. The server then adds contextual information to these basic prompt statements, such as the target audience being review agencies, the required chapter titles in the document, and the document's length range.

[0368] In the sentiment analysis and sentiment feedback control module, the server inputs the user's sentiment state information into the sentiment classification network. This network can be a convolutional neural network (processing text sentiment), a long short-term memory network (processing speech sentiment), or a multimodal attention network (processing both text and images). The server obtains sentiment category labels and sentiment intensity scores through this network. Based on these values, the server then uses rules to map them into several generation parameters, such as parameters for level of detail, weights for reassuring tone, and weights for emphasizing safety. The server injects these parameters into prompt statements, such as adding text instructions like "Please use reassuring expressions and explain the process and risk control measures in detail," forming additional prompt statements. Because the server introduces numerical sentiment parameters and corresponding semantic instructions into the prompt statements, the generative AI model tends to sample words and sentence structures that better match these instructions during internal decoding, thus achieving technical controllability of the generated content.

[0369] In generative AI models, the server can employ a multi-layer network structure based on a self-attention mechanism. During the training phase, the server uses a large number of application document samples and their corresponding structured information as training data. It utilizes the cross-entropy loss function to measure the difference between the generated sequence and the target sequence, and updates the network weights through backpropagation. During training, the server can employ techniques such as learning rate scheduling, gradient pruning, and regularization to avoid overfitting and improve training stability. The server can utilize data augmentation techniques, such as rearranging paragraphs and replacing synonyms in existing documents, to enhance the model's robustness to different text styles and expressions. During the inference phase, the server uses beam search or temperature-controlled sampling strategies to generate the most suitable output text from a probability distribution.

[0370] In the document generation and format conversion module, the server performs post-processing on the text output by the generative AI model. First, the server checks if the document contains all expected chapter titles and key fields. If some fields are missing, it generates additional prompts to request the generative AI model to fill in the missing content. The server then stores the verified document in the database and calls the document conversion component to convert the plain text into an electronic document format according to a pre-defined template, such as fixed page size, title styles, and table layouts. This conversion process can be achieved through a template engine and layout library, enabling the server-generated document to be directly parsed and archived by external review devices.

[0371] In its external communication module, the server interacts with external review devices or support system providers via a network interface. The server can employ secure communication protocols to encrypt and transmit documents and their metadata, and update the application status field in the database upon receiving feedback from external systems. When necessary, the server sends notifications to the terminal, which then displays status information such as "Accepted," "Under Review," or "Request for Correction" on its display device.

[0372] In this invention, the terminal is primarily responsible for user interaction and information presentation. The terminal's user interface can provide a natural language input box, an emotion state selection control, and a document preview and editing area. The user inputs project description text into the terminal, for example: "We plan to install a new 500kW solar power system on the factory roof and hope to obtain relevant subsidies." The terminal sends the text along with emotional selection information (such as "anxious" or "stressed") to the server. After receiving the application document returned by the server, the terminal displays it in an editable format, allowing the user to directly modify the wording or add technical details. After making the modifications, the user resubmits the revised document to the server via the terminal, so that the server generates a second application document and sends it to an external review unit.

[0373] In a specific example, the server receives the following prompt statement as a basis: "Please generate a grant application for the factory rooftop solar power system project based on the following information."

[0374] Project Name: Factory Rooftop Solar Power Generation System Implementation Project Project Overview: Install a 500kW solar power system on the factory roof. The budget is approximately 30 million yuan, and the construction period is one year. The goal is to reduce carbon emissions and increase energy self-sufficiency.

[0375] Corresponding system: Support system for the introduction of renewable energy equipment Please generate a text document following the format of an official application, including sections on project background, implementation details, budget breakdown, expected outcomes, risks, and countermeasures. When the user's emotional state is "anxious", the server further constructs additional prompts: "Users are currently feeling uneasy. Please maintain a formal business style while using a reassuring tone to explain in detail the application process, key review points, and the support measures this policy provides for project risks." The server merges the basic prompts and additional prompts into the generative artificial intelligence model. Through internal attention mechanisms and conditional decoding processes, it adjusts the tone and content distribution of the output text, so that the generated document not only meets the requirements of the system, but also optimizes the expression for the user's psychological state.

[0376] Through the above-described specific structure and processing flow, the server of the present invention achieves the following technical effects: The server reduces fuzzy searches based on raw text and improves the accuracy of matching supported systems by using natural language parsing and structured information generation. Through a vectorized scoring network, the server unifies the modeling of multi-dimensional project features and system features, achieving efficient matching under complex conditions. By constructing prompts and injecting sentiment parameters, the server explicitly controls the decoding behavior of the generative AI model, reducing the generation of irrelevant content and repeated recalculations, thereby lowering computational resource consumption and communication round trips. Through a unified data structure and modular processing links, the server enables the entire application document generation and submission process to be executed efficiently and traceably within the computer, achieving a substantial technological improvement compared to traditional manual input and rule-based concatenation methods.

[0377] In other implementations, the server can employ different network structures, such as using a bidirectional recurrent network as an encoder or a graph neural network to model the similarity relationships between regulations, thereby further improving the accuracy of regulation recommendations. The server can also adjust the loss function according to different application scenarios; for example, it can introduce coverage loss when training generative AI models to reduce the probability of missing key fields in generated documents. The server can also employ an incremental learning mechanism. When external review devices provide feedback such as "pass," "reject," or "correction," the server uses these results along with the current project features and generated documents as new training samples, continuously optimizing the matching and generation models through periodic retraining.

[0378] Therefore, by specifically designing the internal module structure, data structure, feature representation, network architecture, and prompt statement construction algorithm of the server, this invention enables the computer system to perform complex numerical calculations and data processing in a high-precision and high-efficiency manner when handling tasks such as project description, support system selection, and application document generation. This not only replaces some manual work, but also achieves performance improvement and resource utilization optimization at the technical levels of natural language understanding, matching operation, and text generation.

[0379] use Figure 14 The processing flow is explained.

[0380] Step 1: Users input event information and emotional information on the terminal. Users open the application interface in the terminal, enter relevant information in the text input box, such as project name, project background, budget, period and implementation content, and select their current emotional state in the emotion selection control, such as "anxious", "stressed" or "reassuring".

[0381] Input: Original event text, sentiment selection (or sentiment-related data such as voice / image), user identifier.

[0382] The terminal performs initial encapsulation of the above input, combining the text, sentiment options, and user identifier into structured message data, such as a data object containing key-value pairs of fields. The terminal does not perform complex parsing of the text content locally; it only performs character encoding standardization and length checks.

[0383] Output: Encapsulated message data. The terminal sends this message data to the server's designated receiving interface via the communication interface.

[0384] Step 2: The server receives and stores the raw input data. The server receives message data from the terminal through the network interface, parses the data object, and separates the event text field, sentiment field, and user identifier field.

[0385] Input: Encapsulated message data.

[0386] The server writes the original text, sentiment status, and user identifier into a new record in the database, assigning a unique item identifier to that record. The server also records the receiving timestamp and source terminal information. In this step, the server primarily performs data parsing and persistent writing operations; it does not alter the text content, only performing format normalization.

[0387] Output: Database record identifiers containing the original event text and sentiment information.

[0388] Step 3: The server performs text preprocessing and natural language parsing. The server reads the original item text corresponding to the item identifier from the database, and uses a natural language processing library to perform preprocessing operations on the text, such as sentence segmentation, removal of invalid symbols, and unit standardization.

[0389] Input: Original event text.

[0390] The server performs word segmentation, part-of-speech tagging, and named entity recognition on the preprocessed text, and maps words to embedding vectors in an internal tensor structure. The server then performs matrix multiplication, nonlinear transformations, and attention weight calculations on these vectors through a multi-layer network to obtain the contextual semantic representation of each word and each entity.

[0391] Data processing: converting character sequences into word sequences, converting word sequences into vector sequences, and generating labeled entities and syntactic structures from vector sequences.

[0392] Output: Parsing results data including word list, entity list, syntactic dependency relations, and corresponding vector representations.

[0393] Step 4: The server extracts structured information and key information. Based on the parsing results, the server applies rules and a trained classification model to extract fields such as item type, purpose, scale, budget, period, and target domain from the entity list and syntactic structure.

[0394] Input: Parsing result data (word list, entity list, syntactic structure, vector representation).

[0395] The server performs the following data operations: The server inputs sentence vectors representing the entire text into a classification network, outputting item type and target domain labels; the server performs unit normalization on monetary amounts, converting amounts in different formats into a unified value; the server parses time entities, converting "half a year," "one year," etc., into specific months. The server also selects several high-weight terms from the text as key information.

[0396] Output: A structured information object containing structured fields (item type, budget, period, target area, etc.) and a list of key information.

[0397] Step 5: The server retrieves candidate support systems and calculates a suitability score. The server reads characteristic data of all available supporting systems from the system information storage device, including applicable areas, budget scope, target type, and region.

[0398] Input: Structured information objects and institutional feature data sets.

[0399] The server first performs filtering operations based on conditions such as domain and budget limits to obtain a preliminary candidate set. Then, the server inputs the feature vector of the project and the feature vector of each candidate system into a matching network or scoring function, and calculates the fit score of each candidate through vector dot product, fully connected layers and activation functions.

[0400] Data processing: Convert symbolic rules into numerical features, compare multidimensional features, and calculate continuous scores.

[0401] Output: A list of candidate support systems with compatibility scores.

[0402] Step 6: The server selects the optimal support system and generates system requirement information. The server sorts the candidate support system list according to the suitability score and selects the one or more with the highest score as the optimal support system.

[0403] Input: A list of candidate support systems with fit scores.

[0404] The server reads the policy requirements corresponding to the optimal support policy from the policy information storage device, including required fields, a list of required materials, and text structure requirements. The server assembles these policy requirements with the project's structured information into a unified data structure, providing a complete context for the generation of subsequent prompt statements.

[0405] Output: A system selection result object containing the optimal support system identifier and its system requirements information.

[0406] Step 7: The server constructs basic prompts for generative artificial intelligence models. The server generates basic prompt statements based on the project's structured information and policy requirements, following a predefined template.

[0407] Input: Structured information objects, system selection result objects.

[0408] The server performs string concatenation and template filling operations, inserting specific field values ​​into placeholder positions within the template. For example, it converts project name, project overview, budget amount, and duration into natural language descriptions, along with the generated requirements. An example prompt statement is: "Please generate a grant application for the factory rooftop solar power system project based on the following information."

[0409] Project Name: Factory Rooftop Solar Power Generation System Implementation Project Project Overview: Install a 500kW solar power system on the factory roof. The budget is approximately 30 million yuan, and the construction period is one year. The goal is to reduce carbon emissions and increase energy self-sufficiency.

[0410] Corresponding system: Support system for the introduction of renewable energy equipment Please generate a text document following the format of an official application, including sections on project background, implementation details, budget breakdown, expected outcomes, risks, and countermeasures. Output: The basic prompt text used to generate the application documents.

[0411] Step 8: The server analyzes the user's emotional state and generates additional prompts. The server reads sentiment status information corresponding to the item identifier from the database. If the user only selects a sentiment tag, the server uses that tag directly. If the user provides voice, images, or additional text, the server inputs this data into the sentiment analysis model for processing.

[0412] Input: Emotional state related information (labels, voice data, image data, or text data).

[0413] The server performs feature extraction and classification operations in the sentiment analysis model. For example, it performs spectral analysis on speech and inputs it into a recurrent network, performs emotional vocabulary statistics on text and inputs it into a convolutional network, and obtains the sentiment type and intensity values ​​from the output layer. The server sets parameters for detail and tone control based on different types and intensities, and generates additional prompts, such as: "Users are currently feeling uneasy. Please maintain a formal business style while using a reassuring tone to explain in detail the application process, key review points, and the support measures this policy provides for project risks." Output: Additional cue text and a set of sentiment parameters used to control the generation behavior.

[0414] Step 9: The server merges the prompt statements and calls a generative artificial intelligence model to generate application documents. The server merges the basic prompt statement with the additional prompt statement to form a complete prompt statement, and at the same time maps the sentiment parameters to model decoding control parameters (such as temperature, length preference or specific tag marking).

[0415] Input: basic prompts, additional prompts, and a set of sentiment parameters.

[0416] The server encodes complete prompts as input sequences into the model's encoder or input embedding layer via a generative AI model interface. Attention weights between words are calculated in a multi-layer self-attention network, and output words are generated progressively in the decoder. The server can set the bundle search width or sampling strategy to balance diversity and accuracy during the generation process.

[0417] Data processing: extensive matrix multiplication, vector weighting, Softmax probability normalization, and sequence decoding.

[0418] Output: An application document containing the project background, implementation details, budget breakdown, and expected results.

[0419] Step 10: The server verifies the generated document and performs format conversion. The server performs content validation on the generated application document text, such as checking whether it contains required chapter titles, budget fields, and policy names.

[0420] Input: The generated application document text.

[0421] The server uses rule-based detection or simple text classification models to identify missing content. If key parts are found to be missing, the server can reconstruct supplementary prompts and then call a generative AI model to generate the missing paragraphs again. Through this cycle, the server ensures that the final document structurally meets the regulatory requirements. After confirming the completeness of the document content, the server combines the text with the document template, performs layout and format conversion calculations, and outputs an electronic document in a format that meets the requirements of external review devices.

[0422] Output: A complete and formatted electronic application document.

[0423] Step 11: The server sends documents to the terminal for the user to confirm and modify. The server will send the generated and verified application document text to the terminal, which may also include a policy description and submission status information.

[0424] Input: Application document text, project identifier, and policy description data.

[0425] The terminal displays the application document in an editable view on the display device, presenting each chapter in sections and allowing users to directly modify paragraph content or add supplementary explanations. After the user completes editing on the terminal, they click to confirm, and the terminal sends the modified document text along with the project identifier back to the server.

[0426] Output: The user-revised application document text.

[0427] Step 12: The server generates the final application documents and submits them to the external device. The server receives the revised application documents from the user, compares them with the previously generated text, and records the user's modifications for subsequent model optimization or auditing.

[0428] Input: User-revised application document text, original generated document text.

[0429] The server generates a second application document based on the revised text and performs another format conversion to ensure that this version meets the electronic submission format requirements. The server then sends the second application document to an external examination device or support system provider via a communication module, and records the returned receipt result or acceptance number. The server then writes the submission result to the database and sends status update information to the terminal.

[0430] Output: Status record of the final application document that has been successfully submitted and has an acceptance number.

[0431] The specific processing unit 290 sends the result of the specific processing to the smart device 14. In the smart device 14, the control unit 46A causes the output device 40 to output the result of the specific processing. The microphone 38B acquires sound representing user input regarding the result of the specific processing. The control unit 46A sends the sound data representing user input acquired by the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the sound data.

[0432] Data generation model 58 is a so-called generative AI (Artificial Intelligence). Examples of data generation models 58 include ChatGPT (registered trademark) (accessible via the internet (URL: https: / / openai.com / blog / chatgpt)). Data generation model 58 is obtained through deep learning on a neural network. Input to data generation model 58 are prompt words containing instructions, and inference data such as sound data representing sound, text data representing text, and image data representing images (e.g., still image data or animation data). Data generation model 58 infers from the input inference data based on the instructions represented by the prompt words and outputs the inference result in one or more data forms, such as sound data, text data, and image data. Data generation model 58 includes, for example, text generation AI, image generation AI, and multimodal generation AI. Here, inference refers to, for example, analysis, classification, prediction, and / or induction. The specific processing unit 290 performs the aforementioned specific processing while using data generation model 58. The data generation model 58 can also be a model finely tuned to output inference results from prompts that do not contain instructions. In this case, the data generation model 58 can output inference results based on prompts that do not contain instructions. The data processing apparatus 12, etc., includes various data generation models 58, including AI other than the generation AI. AI other than the generation AI can be, for example, linear regression, logistic regression, decision trees, random forests, support vector machines (SVM), k-means clustering, convolutional neural networks (CNN), recurrent neural networks (RNN), generative adversarial networks (GAN), or Naive Bayes, and can perform various processes, but is not limited to this example. Furthermore, the AI ​​can also be an AI agent. Furthermore, when the processing of the above-mentioned parts is performed by AI, the processing can be performed partially or entirely by AI, but is not limited to this example. Furthermore, the processing performed by the AI ​​including the generation AI can be replaced by processing in the rule base, and the processing in the rule base can also be replaced by processing performed by the AI ​​including the generation AI.

[0433] Furthermore, the processing of the aforementioned data processing system 10 is performed by the specific processing unit 290 of the data processing device 12 or the control unit 46A of the smart device 14, but it can also be performed by both the specific processing unit 290 of the data processing device 12 and the control unit 46A of the smart device 14. Additionally, the specific processing unit 290 of the data processing device 12 acquires or collects information required for processing from the smart device 14 or external devices, and the smart device 14 acquires or collects information required for processing from the data processing device 12 or external devices.

[0434] For example, the collection unit is implemented by the control unit 46A of the smart device 14 or the specific processing unit 290 of the data processing device 12. For example, the acquisition unit uses the camera 42 or communication I / F 44 of the smart device 14 to acquire step data, which is then processed by the specific processing unit 290 of the data processing device 12. For example, the analysis unit is implemented by the specific processing unit 290 of the data processing device 12, which analyzes the data from the collection unit and the acquisition unit. For example, the generation unit is implemented by the specific processing unit 290 of the data processing device 12, which uses a generation AI to generate a menu. For example, the serving unit is implemented by the output device 40 of the smart device 14 or the specific processing unit 290 of the data processing device 12, which provides the generated menu to the user. The correspondence between each unit and the device or control unit is not limited to the above examples and various changes can be made.

[0435] In the above embodiments, examples of specific processing by the data processing device 12 are given, but the technology disclosed herein is not limited to this, and specific processing may also be performed by the smart device 14.

[0436] Second Implementation Method Figure 3 An example of the configuration of the data processing system 210 according to the second embodiment is shown.

[0437] like Figure 3 As shown, the data processing system 210 includes a data processing device 12 and smart glasses 214. A server can be cited as an example of the data processing device 12.

[0438] The data processing apparatus 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" as understood in this disclosure. The computer 22 includes a processor 28, RAM 30, and memory 32. The processor 28, RAM 30, and memory 32 are connected to a bus 34. Furthermore, the database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0439] The smart glasses 214 include a computer 36, a microphone 238, a speaker 240, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, RAM 48, and memory 50. The processor 46, RAM 48, and memory 50 are connected to a bus 52. Furthermore, the microphone 238, speaker 240, camera 42, and communication I / F 44 are also connected to the bus 52.

[0440] Microphone 238 receives instructions from user 20 by receiving sounds emitted by user 20. Microphone 238 captures sounds emitted by user 20 and converts the captured sounds into sound data, which is then output to processor 46. Speaker 240 outputs sound according to instructions from processor 46.

[0441] Camera 42 is a small digital camera equipped with an optical system such as a lens, aperture and shutter, and imaging elements such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, which captures images of the user 20's surroundings (e.g., the field of view defined by an angle equivalent to the field of vision of an average healthy person).

[0442] Communication I / F44 is connected to network 54. Communication I / F44 and 26 are responsible for the transmission and reception of various information between processor 46 and processor 28 via network 54. The transmission and reception of various information between processor 46 and processor 28 using communication I / F44 and 26 is performed in a secure state.

[0443] Figure 4 This illustrates an example of the main functions of the data processing device 12 and the smart glasses 214. For example... Figure 4 As shown, in the data processing device 12, specific processing is performed by the processor 28. The specific processing program 56 is stored in the memory 32.

[0444] The specific processing program 56 is an example of a "program" involved in the technology of this disclosure. The processor 28 reads the specific processing program 56 from the memory 32 and executes the read specific processing program 56 on the RAM 30. Specific processing is implemented by the processor 28 operating as a specific processing unit 290 according to the specific processing program 56 executed on the RAM 30.

[0445] The memory 32 stores a data generation model 58 and an emotion-specific model 59. The data generation model 58 and the emotion-specific model 59 are used by the specific processing unit 290. The specific processing unit 290 can use the emotion-specific model 59 to infer the user's emotions and perform specific processing based on the user's emotions. In the emotion inference function (emotion-specific function) using the emotion-specific model 59, various inferences and predictions related to the user's emotions are performed, including inferences and predictions of the user's emotions, but this is not limited to this example. Furthermore, emotion inference and prediction may also include, for example, emotion analysis (parsing).

[0446] In the smart glasses 214, the processor 46 performs reception and output processing. The memory 50 stores the reception and output program 60. The processor 46 reads the reception and output program 60 from the memory 50 and executes the read reception and output program 60 on the RAM 48. The reception and output processing is implemented by the processor 46 operating as a control unit 46A according to the reception and output program 60 executed on the RAM 48. Furthermore, the smart glasses 214 has the same data generation model and emotion-specific model as the data generation model 58 and the emotion-specific model 59, and these models can also be used to perform the same processing as the specific processing unit 290.

[0447] Next, the specific processing of the specific processing unit 290 of the data processing device 12 will be described. Each part of the system described below is implemented by the data processing device 12 and the smart glasses 214. In the following description, the data processing device 12 will be referred to as the "server" and the smart glasses 214 will be referred to as the "terminal".

[0448] Example 1 The process is the same as that of the specific process described in Embodiment 1 in the first embodiment above, so the description is omitted.

[0449] Application Example 1 The process is the same as that in the specific processing described in Application Example 1 of the first embodiment above, so the description is omitted.

[0450] Example 2 The process is the same as that of the specific process in Embodiment 2 described in the first embodiment above, so the description is omitted.

[0451] Application Example 2 The process is the same as that in the specific processing described in Application Example 2 of the first embodiment above, so the description is omitted.

[0452] The specific processing unit 290 sends the result of the specific processing to the smart glasses 214. In the smart glasses 214, the control unit 46A outputs the result of the specific processing to the speaker 240. The microphone 238 acquires sound input representing the user's input regarding the result of the specific processing. The control unit 46A sends the sound data representing the user's input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the sound data.

[0453] Data generation model 58 is a so-called generative AI (Artificial Intelligence). Examples of data generation models 58 include ChatGPT (registered trademark) (accessible via the internet (URL: https: / / openai.com / blog / chatgpt)). Data generation model 58 is obtained through deep learning on a neural network. Input to data generation model 58 are prompt words containing instructions, and inference data such as sound data representing sound, text data representing text, and image data representing images (e.g., still image data or animation data). Data generation model 58 infers from the input inference data based on the instructions represented by the prompt words and outputs the inference result in one or more data forms, such as sound data, text data, and image data. Data generation model 58 includes, for example, text generation AI, image generation AI, and multimodal generation AI. Here, inference refers to, for example, analysis, classification, prediction, and / or induction. The specific processing unit 290 performs the aforementioned specific processing while using data generation model 58. The data generation model 58 can also be a model finely tuned to output inference results from prompts that do not contain instructions. In this case, the data generation model 58 can output inference results based on prompts that do not contain instructions. The data processing apparatus 12, etc., includes various data generation models 58, including AI other than the generation AI. AI other than the generation AI can be, for example, linear regression, logistic regression, decision trees, random forests, support vector machines (SVM), k-means clustering, convolutional neural networks (CNN), recurrent neural networks (RNN), generative adversarial networks (GAN), or Naive Bayes, and can perform various processes, but is not limited to this example. Furthermore, the AI ​​can also be an AI agent. Furthermore, when the processing of the above-mentioned parts is performed by AI, the processing can be performed partially or entirely by AI, but is not limited to this example. Furthermore, the processing performed by the AI ​​including the generation AI can be replaced by processing in the rule base, and the processing in the rule base can also be replaced by processing performed by the AI ​​including the generation AI.

[0454] Furthermore, the processing of the aforementioned data processing system 10 is performed by the specific processing unit 290 of the data processing device 12 or the control unit 46A of the smart glasses 214, but it can also be performed by both the specific processing unit 290 of the data processing device 12 and the control unit 46A of the smart glasses 214. Additionally, the specific processing unit 290 of the data processing device 12 acquires or collects information required for processing from the smart glasses 214 or external devices, and the smart glasses 214 acquires or collects information required for processing from the data processing device 12 or external devices.

[0455] For example, the collection unit is implemented by the control unit 46A of the smart glasses 214 or the specific processing unit 290 of the data processing device 12. For example, the acquisition unit uses the camera 42 or communication I / F 44 of the smart glasses 214 to acquire step data, which is then processed by the specific processing unit 290 of the data processing device 12. For example, the analysis unit is implemented by the specific processing unit 290 of the data processing device 12, which analyzes the data from the collection unit and the acquisition unit. For example, the generation unit is implemented by the specific processing unit 290 of the data processing device 12, which uses a generation AI to generate a menu. For example, the serving unit is implemented by the speaker 240 of the smart glasses 214 or the specific processing unit 290 of the data processing device 12, which provides the generated menu to the user. The correspondence between each unit and the device or control unit is not limited to the above examples and various changes can be made.

[0456] In the above embodiments, examples of specific processing by the data processing device 12 are given, but the technology disclosed herein is not limited to this, and specific processing may also be performed by the smart glasses 214.

[0457] Third Implementation Method Figure 5 An example of the configuration of the data processing system 310 according to the third embodiment is shown.

[0458] like Figure 5 As shown, the data processing system 310 includes a data processing device 12 and a head-mounted terminal 314. A server can be cited as an example of the data processing device 12.

[0459] The data processing apparatus 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" as understood in this disclosure. The computer 22 includes a processor 28, RAM 30, and memory 32. The processor 28, RAM 30, and memory 32 are connected to a bus 34. Furthermore, the database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0460] The head-mounted terminal 314 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a display 343. The computer 36 includes a processor 46, RAM 48, and memory 50. The processor 46, RAM 48, and memory 50 are connected to a bus 52. Furthermore, the microphone 238, speaker 240, camera 42, display 343, and communication I / F 44 are also connected to the bus 52.

[0461] Microphone 238 receives instructions from user 20 by receiving sounds emitted by user 20. Microphone 238 captures sounds emitted by user 20 and converts the captured sounds into sound data, which is then output to processor 46. Speaker 240 outputs sound according to instructions from processor 46.

[0462] Camera 42 is a small digital camera equipped with an optical system such as a lens, aperture and shutter, and imaging elements such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, which captures images of the user 20's surroundings (e.g., the field of view defined by an angle equivalent to the field of vision of an average healthy person).

[0463] Communication I / F44 is connected to network 54. Communication I / F44 and 26 are responsible for the transmission and reception of various information between processor 46 and processor 28 via network 54. The transmission and reception of various information between processor 46 and processor 28 using communication I / F44 and 26 is performed in a secure state.

[0464] Figure 6 This illustrates an example of the main functions of the data processing device 12 and the head-mounted terminal 314. For example... Figure 6 As shown, in the data processing device 12, specific processing is performed by the processor 28. The specific processing program 56 is stored in the memory 32.

[0465] The specific processing program 56 is an example of a "program" involved in the technology of this disclosure. The processor 28 reads the specific processing program 56 from the memory 32 and executes the read specific processing program 56 on the RAM 30. Specific processing is implemented by the processor 28 operating as a specific processing unit 290 according to the specific processing program 56 executed on the RAM 30.

[0466] The memory 32 stores the data generation model 58 and the emotion-specific model 59. The data generation model 58 and the emotion-specific model 59 are used by the specific processing unit 290.

[0467] In the head-mounted terminal 314, the processor 46 performs the acceptance / output processing. The memory 50 stores the acceptance / output program 60. The processor 46 reads the acceptance / output program 60 from the memory 50 and executes the read acceptance / output program 60 on the RAM 48. The acceptance / output processing is implemented by the processor 46 operating as a control unit 46A according to the acceptance / output program 60 executed on the RAM 48.

[0468] Next, the specific processing of the specific processing unit 290 of the data processing device 12 will be described. Each part of the system described below is implemented by the data processing device 12 and the head-mounted terminal 314. In the following description, the data processing device 12 will be referred to as the "server" and the head-mounted terminal 314 will be referred to as the "terminal".

[0469] Example 1 The process is the same as that of the specific process described in Embodiment 1 in the first embodiment above, so the description is omitted.

[0470] Application Example 1 The process is the same as that in the specific processing described in Application Example 1 of the first embodiment above, so the description is omitted.

[0471] Example 2 The process is the same as that of the specific process in Embodiment 2 described in the first embodiment above, so the description is omitted.

[0472] Application Example 2 The process is the same as that in the specific processing described in Application Example 2 of the first embodiment above, so the description is omitted.

[0473] The specific processing unit 290 sends the result of the specific processing to the head-mounted terminal 314. In the head-mounted terminal 314, the control unit 46A outputs the result of the specific processing to the speaker 240 and the display 343. The microphone 238 acquires sound input representing the user's input regarding the result of the specific processing. The control unit 46A sends the sound data representing the user's input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the sound data.

[0474] Data generation model 58 is a so-called generative AI (Artificial Intelligence). Examples of data generation models 58 include ChatGPT (registered trademark) (accessible via the internet (URL: https: / / openai.com / blog / chatgpt)). Data generation model 58 is obtained through deep learning on a neural network. Input to data generation model 58 includes prompt words containing instructions, and inference data such as sound data representing sound, text data representing text, and image data representing images (e.g., still image data or animation data). Data generation model 58 infers the input inference data based on the instructions represented by the prompt words and outputs the inference result in one or more data forms such as sound data, text data, and image data. Data generation model 58 includes, for example, text generation AI, image generation AI, and multimodal generation AI. Here, inference refers to, for example, analysis, classification, prediction, and / or induction. The specific processing unit 290 performs the aforementioned specific processing while using data generation model 58. The data generation model 58 can also be a model finely tuned to output inference results from prompts that do not contain instructions. In this case, the data generation model 58 can output inference results based on prompts that do not contain instructions. The data processing apparatus 12, etc., includes various data generation models 58, including AI other than the generation AI. AI other than the generation AI can be, for example, linear regression, logistic regression, decision trees, random forests, support vector machines (SVM), k-means clustering, convolutional neural networks (CNN), recurrent neural networks (RNN), generative adversarial networks (GAN), or Naive Bayes, and can perform various processes, but is not limited to this example. Furthermore, the AI ​​can also be an AI agent. Furthermore, when the processing of the above-mentioned parts is performed by AI, the processing can be performed partially or entirely by AI, but is not limited to this example. Furthermore, the processing performed by the AI ​​including the generation AI can be replaced by processing in the rule base, and the processing in the rule base can also be replaced by processing performed by the AI ​​including the generation AI.

[0475] Furthermore, the processing of the aforementioned data processing system 10 is performed by the specific processing unit 290 of the data processing device 12 or the control unit 46A of the head-mounted terminal 314, but it can also be performed by both the specific processing unit 290 of the data processing device 12 and the control unit 46A of the head-mounted terminal 314. Additionally, the specific processing unit 290 of the data processing device 12 acquires or collects information required for processing from the head-mounted terminal 314 or external devices, and the head-mounted terminal 314 acquires or collects information required for processing from the data processing device 12 or external devices.

[0476] For example, the collection unit is implemented by the control unit 46A of the head-mounted terminal 314 or the specific processing unit 290 of the data processing device 12. For example, the acquisition unit uses the camera 42 or communication I / F 44 of the head-mounted terminal 314 to acquire step data, which is then processed by the specific processing unit 290 of the data processing device 12. For example, the analysis unit is implemented by the specific processing unit 290 of the data processing device 12 to analyze the data from the collection unit and the acquisition unit. For example, the generation unit is implemented by the specific processing unit 290 of the data processing device 12 to generate a menu using a generation AI. For example, the serving unit is implemented by the speaker 240 and display 343 of the head-mounted terminal 314 or the specific processing unit 290 of the data processing device 12 to provide the generated menu to the user. The correspondence between each unit and the device or control unit is not limited to the above examples and various changes can be made.

[0477] In the above embodiments, examples of specific processing by the data processing device 12 are given, but the technology disclosed herein is not limited to this, and specific processing may also be performed by the head-mounted terminal 314.

[0478] Fourth Implementation Method Figure 7 An example of the configuration of the data processing system 410 according to the fourth embodiment is shown.

[0479] like Figure 7 As shown, the data processing system 410 includes a data processing device 12 and a robot 414. A server can be cited as an example of the data processing device 12.

[0480] The data processing apparatus 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" as understood in this disclosure. The computer 22 includes a processor 28, RAM 30, and memory 32. The processor 28, RAM 30, and memory 32 are connected to a bus 34. Furthermore, the database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0481] Robot 414 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a controlled object 443. The computer 36 includes a processor 46, RAM 48, and memory 50. The processor 46, RAM 48, and memory 50 are connected to a bus 52. Furthermore, the microphone 238, speaker 240, camera 42, controlled object 443, and communication I / F 44 are also connected to the bus 52.

[0482] Microphone 238 receives instructions from user 20 by receiving sounds emitted by user 20. Microphone 238 captures sounds emitted by user 20 and converts the captured sounds into sound data, which is then output to processor 46. Speaker 240 outputs sound according to instructions from processor 46.

[0483] Camera 42 is a small digital camera equipped with an optical system such as a lens, aperture and shutter, and imaging elements such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, to photograph the area around robot 414 (e.g., the field of view defined by a perspective equivalent to the field of vision of an average healthy person).

[0484] Communication I / F44 is connected to network 54. Communication I / F44 and 26 are responsible for the transmission and reception of various information between processor 46 and processor 28 via network 54. The transmission and reception of various information between processor 46 and processor 28 using communication I / F44 and 26 is performed in a secure state.

[0485] The controlled object 443 includes a display device, LEDs (light-emitting diodes) for the eyes, and motors for driving the arms, hands, and feet. The posture or movement of the robot 414 is controlled by controlling the motors in the arms, hands, and feet. Some of the robot 414's emotions can be expressed by controlling these motors. In addition, the robot 414's facial expressions can also be expressed by controlling the illumination state of the LEDs in its eyes.

[0486] Figure 8 This illustrates an example of the main functions of the data processing device 12 and the robot 414. For example... Figure 8 As shown, in the data processing device 12, specific processing is performed by the processor 28. The specific processing program 56 is stored in the memory 32.

[0487] The specific processing program 56 is an example of a "program" involved in the technology of this disclosure. The processor 28 reads the specific processing program 56 from the memory 32 and executes the read specific processing program 56 on the RAM 30. Specific processing is implemented by the processor 28 operating as a specific processing unit 290 according to the specific processing program 56 executed on the RAM 30.

[0488] The memory 32 stores the data generation model 58 and the emotion-specific model 59. The data generation model 58 and the emotion-specific model 59 are used by the specific processing unit 290.

[0489] In robot 414, the processor 46 performs the acceptance and output processing. The memory 50 stores the acceptance and output program 60. The processor 46 reads the acceptance and output program 60 from the memory 50 and executes the read acceptance and output program 60 on RAM 48. The acceptance and output processing is implemented by the processor 46 acting as the control unit 46A according to the acceptance and output program 60 executed on RAM 48.

[0490] Next, the specific processing of the specific processing unit 290 of the data processing device 12 will be described. Each part of the system described below is implemented by the data processing device 12 and the robot 414. In the following description, the data processing device 12 will be referred to as the "server" and the robot 414 will be referred to as the "terminal".

[0491] Example 1 The process is the same as that of the specific process described in Embodiment 1 in the first embodiment above, so the description is omitted.

[0492] Application Example 1 The process is the same as that in the specific processing described in Application Example 1 of the first embodiment above, so the description is omitted.

[0493] Example 2 The process is the same as that of the specific process in Embodiment 2 described in the first embodiment above, so the description is omitted.

[0494] Application Example 2 The process is the same as that in the specific processing described in Application Example 2 of the first embodiment above, so the description is omitted.

[0495] The specific processing unit 290 sends the result of the specific processing to the robot 414. In the robot 414, the control unit 46A outputs the result of the specific processing to the speaker 240 and the controlled object 443. The microphone 238 acquires sound input representing the result of the specific processing. The control unit 46A sends the sound data representing the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the sound data.

[0496] Data generation model 58 is a so-called generative AI (Artificial Intelligence). Examples of data generation models 58 include ChatGPT (registered trademark) (accessible via the internet (URL: https: / / openai.com / blog / chatgpt)). Data generation model 58 is obtained through deep learning on a neural network. Input to data generation model 58 are prompt words containing instructions, and inference data such as sound data representing sound, text data representing text, and image data representing images (e.g., still image data or animation data). Data generation model 58 infers from the input inference data based on the instructions represented by the prompt words and outputs the inference result in one or more data forms, such as sound data, text data, and image data. Data generation model 58 includes, for example, text generation AI, image generation AI, and multimodal generation AI. Here, inference refers to, for example, analysis, classification, prediction, and / or induction. The specific processing unit 290 performs the aforementioned specific processing while using data generation model 58. The data generation model 58 can also be a model finely tuned to output inference results from prompts that do not contain instructions. In this case, the data generation model 58 can output inference results based on prompts that do not contain instructions. The data processing apparatus 12, etc., includes various data generation models 58, including AI other than the generation AI. AI other than the generation AI can be, for example, linear regression, logistic regression, decision trees, random forests, support vector machines (SVM), k-means clustering, convolutional neural networks (CNN), recurrent neural networks (RNN), generative adversarial networks (GAN), or Naive Bayes, and can perform various processes, but is not limited to this example. Furthermore, the AI ​​can also be an AI agent. Furthermore, when the processing of the above-mentioned parts is performed by AI, the processing can be performed partially or entirely by AI, but is not limited to this example. Furthermore, the processing performed by the AI ​​including the generation AI can be replaced by processing in the rule base, and the processing in the rule base can also be replaced by processing performed by the AI ​​including the generation AI.

[0497] Furthermore, the processing of the aforementioned data processing system 10 is performed by the specific processing unit 290 of the data processing device 12 or the control unit 46A of the robot 414, but it can also be performed by both the specific processing unit 290 of the data processing device 12 and the control unit 46A of the robot 414. Additionally, the specific processing unit 290 of the data processing device 12 acquires or collects information required for processing from the robot 414 or external devices, and the robot 414 acquires or collects information required for processing from the data processing device 12 or external devices.

[0498] For example, the collection unit is implemented by the control unit 46A of the robot 414 or the specific processing unit 290 of the data processing device 12. For example, the acquisition unit uses the camera 42 or communication I / F 44 of the robot 414 to acquire step data, which is then processed by the specific processing unit 290 of the data processing device 12. For example, the analysis unit is implemented by the specific processing unit 290 of the data processing device 12, which analyzes the data from the collection unit and the acquisition unit. For example, the generation unit is implemented by the specific processing unit 290 of the data processing device 12, which uses a generation AI to generate a menu. For example, the serving unit is implemented by the speaker 240 of the robot 414 and the control object 443 or the specific processing unit 290 of the data processing device 12, which provides the generated menu to the user. The correspondence between each unit and the device or control unit is not limited to the above examples and various changes can be made.

[0499] In the above embodiments, examples of specific processing by the data processing device 12 are given, but the technology disclosed herein is not limited to this, and specific processing may also be performed by the robot 414.

[0500] Furthermore, the emotion-specific model 59, acting as an emotion engine, can determine a user's emotion based on a specific mapping. Specifically, the emotion-specific model 59 can determine a user's emotion based on an emotion graph that serves as a specific mapping (see...). Figure 9 The emotion-specific model 59 can also determine the robot's emotion, and the specific processing unit 290 performs specific processing based on the robot's emotions.

[0501] Figure 9 This is a diagram representing an emotion map 400 that maps multiple emotions. In the emotion map 400, emotions are arranged in concentric circles radiating from the center. The closer to the center of the concentric circles, the more primitive the emotion is. On the outer side of the concentric circles, emotions representing states or behaviors arising from mood are arranged. Emotions are concepts that include feelings and mental states. Emotions generated by reactions occurring in the brain are arranged roughly to the left of the concentric circles. Emotions derived from situational judgments are arranged roughly to the right of the concentric circles. Emotions generated by reactions occurring in the brain and derived from situational judgments are arranged roughly above and below the concentric circles. Furthermore, "pleasant" emotions are arranged above the concentric circles, and "unpleasant" emotions are arranged below them. Thus, in the emotion map 400, multiple emotions are mapped based on the structure that generates emotions, and emotions that are likely to occur simultaneously are mapped close to each other.

[0502] These emotions are distributed at the three o'clock position of the emotion map 400, typically fluctuating between peace and anxiety. In the right half of the emotion map 400, situational awareness dominates over internal sensation, thus resulting in an impression of calm.

[0503] The inner side of the emotion map 400 represents the inner state, while the outer side represents behavior. Therefore, the further outward you are from the emotion map 400, the more visible the emotion becomes (manifested in behavior).

[0504] Here, human emotions are based on various balances such as posture and blood sugar levels. When these balances deviate from an ideal state, it indicates an unpleasant state; when they approach the ideal state, it indicates a pleasant state. Emotions in robots, cars, motorcycles, etc., can also be created in the following way: based on various balances such as posture and remaining battery power, when these balances deviate from an ideal state, it indicates an unpleasant state; when they approach the ideal state, it indicates a pleasant state. Emotion maps can be generated, for example, based on Dr. Mitsuyoshi's emotion map (Research on a Brain Physiological Signal Analysis System for Voice Emotion Recognition and Emotion, Tokushima University, Doctoral Dissertation: https: / / ci.nii.ac.jp / naid / 500000375379). In the left half of the emotion map, emotions belonging to the sensory-dominated region, called "response," are arranged. Furthermore, in the right half of the emotion map, emotions belonging to the situational cognition-dominated region, called "situation," are arranged.

[0505] In the emotion map, two types of emotions that promote learning are defined. One is a negative emotion on the situational side, in the middle or peripheral region of "repentance" or "reflection." This occurs when the robot experiences negative emotions such as "I don't want to experience this feeling again" or "I don't want to be blamed again." The other is a positive emotion on the response side, near the "desire" region. This occurs when there are positive feelings such as "wanting more" or "wanting to know more."

[0506] The emotion-specific model 59 inputs user input into a pre-trained neural network to obtain emotion values ​​representing each emotion shown in the emotion map 400, thereby determining the user's emotion. This neural network is pre-trained based on multiple training data sets, which are combinations of user input and emotion values ​​representing each emotion shown in the emotion map 400. Furthermore, this neural network... Figure 10 As shown in the sentiment graph 900, it was trained in a way that sentiments that are configured close to each other have similar values. Figure 10 The text shows examples of emotions such as "peace of mind", "stability", and "reassurance" that have similar emotion values.

[0507] The above description focuses on the functions of the data processing device 12, but the system of this disclosure is not necessarily installed on a server. The system of this disclosure can also be installed as a general information processing system. This disclosure can also be installed, for example, as a software program running on a personal computer, an application running on a smartphone, etc. The method of this disclosure can also be provided to users in the form of SaaS (Software as a Service).

[0508] In the above embodiments, an example of a specific process being performed by a single computer 22 is given. However, the technology disclosed herein is not limited to this, and the specific process can also be distributed among multiple computers, including computer 22. For example, the data generation model 58 can be located on an external device of the data processing apparatus 12, where data is generated based on the input data.

[0509] In the above embodiments, examples of storing a specific processing program 56 in the memory 32 have been described, but the technology disclosed herein is not limited thereto. For example, the specific processing program 56 may also be stored in a portable computer-readable non-transitory storage medium such as a USB (Universal Serial Bus) memory. The specific processing program 56 stored in the non-transitory storage medium is installed into the computer 22 of the data processing device 12. The processor 28 executes specific processing according to the specific processing program 56.

[0510] Alternatively, a specific processing program 56 may be pre-stored in a storage device such as a server connected to the data processing device 12 via the network 54, and the specific processing program 56 may be downloaded and installed on the computer 22 according to the requirements of the data processing device 12.

[0511] In addition, it is not necessary to store all the specific processing program 56 in the storage device such as the server connected to the data processing device 12 via the network 54 or in the memory 32; a portion of the specific processing program 56 may be stored in advance.

[0512] As hardware resources for performing specific processes, various processors, as shown below, can be used. For example, a CPU can be listed as a processor, which functions as a general-purpose processor that performs specific processes by executing software, i.e., a program. Furthermore, processors can be listed as special-purpose circuits such as FPGAs (Field-Programmable Gate Arrays), PLDs (Programmable Logic Devices), or ASICs (Application-Specific Integrated Circuits), which are processors with circuitry specifically designed to perform specific processes. Each processor has built-in or connected memory, and each processor executes specific processes using that memory.

[0513] The hardware resources for performing a specific process can consist of one of these various processors, or a combination of two or more processors of the same or different types (e.g., a combination of multiple FPGAs, or a combination of a CPU and an FPGA). Alternatively, the hardware resources for performing a specific process can be a single processor.

[0514] As an example of a single processor, there are two approaches: First, a processor is composed of a combination of one or more CPUs and software, which functions as a hardware resource to perform a specific process; second, as represented by a SoC (System-on-a-chip), a processor is used to implement the functionality of the entire system, which includes multiple hardware resources for performing a specific process, using a single IC (Integrated Circuit) chip. In this way, the specific process is implemented by using one or more of the aforementioned processors as hardware resources.

[0515] Furthermore, the hardware architecture of these various processors, more specifically, can utilize circuits that combine semiconductor elements and other circuit components. Moreover, the specific process described above is just one example. Therefore, without departing from the main point, unnecessary steps can certainly be deleted, new steps added, or the processing order changed.

[0516] The descriptions and illustrations above are detailed explanations of a portion of the technology disclosed herein, and are merely one example of the technology disclosed herein. For example, the above descriptions of the structure, function, effect, and results are just one example of the structure, function, effect, and results of a portion of the technology disclosed herein. Therefore, without departing from the spirit of the technology disclosed herein, unnecessary parts may be deleted, new elements added, or replacements may be made to the descriptions and illustrations above. Furthermore, to avoid confusion and facilitate understanding of a portion of the technology disclosed herein, explanations of common technical knowledge that do not require special explanation under the premise of being able to implement the technology disclosed herein have been omitted from the descriptions and illustrations above.

[0517] All documents, patent applications and technical specifications set forth in this specification are incorporated herein by reference to the same extent that each document, patent application and technical specification is specifically and individually described therein and referenced by reference.

[0518] In addition, the following notes are provided in response to the above explanation.

[0519] Example 1 (Note 1) An information processing system, characterized in that it comprises: This is a means of obtaining business content information in natural language form from a user terminal and storing the business content information as structured application input information in a data storage device. A means for dynamically generating prompt statements for a generative artificial intelligence model based on the business content information as input, and calling the generative artificial intelligence model to parse the business content information using natural language processing to obtain parsing results including business classification information, target domain information, target subject information, and region information. This is a means to perform retrieval processing on an information storage device storing information on financial assistance systems based on the analysis results, extract multiple candidate financial assistance systems while considering information on applicable conditions and target domain information, and, when necessary, perform adaptation calculation processing based on evaluation prompts using the generative artificial intelligence model to evaluate the multiple candidate financial assistance systems and determine the optimal financial assistance system. A means for obtaining document style information corresponding to a specific funding assistance system, generating a document generation prompt statement for automatically allocating the parsing results and a portion of the business content information to placeholders in the document style information, and automatically generating application document data based on the document generation prompt statement; Means for converting the application document data into data for display and presenting it to a user interface, and means for receiving additional input information and modification information from the user through the user interface, and updating the application document data and the application input information in the data storage device according to the additional input information and modification information; This is a means of sending the application document data as an electronic application completed to an external processing device or external receiving device based on the updated application document data and the sending destination information associated with the specific funding assistance scheme, and recording the sending result as management information.

[0520] (Note 2) According to the information processing system described in Appendix 1, the system further includes: means for inferring a user's emotional state based on user interaction information obtained from a user terminal and sentiment analysis prompts using the generative artificial intelligence model; and means for generating adjustment information based on the emotional state to adjust the document generation prompts or the content expressed in the application document data, and for dynamically changing the recording style or description level of the application document data based on the adjustment information.

[0521] (Note 3) According to the information processing system described in Appendix 1, the user interface is configured to: output the application document data in the form of an editable area, along with the system requirements information and application deadline information of the specific funding assistance system; acquire user changes to each item to manage the changes by item; generate additional input prompts for the generative artificial intelligence model based on the changes; and regenerate or complete at least a portion of the application document data using the output of the generative artificial intelligence model.

[0522] Application Example 1 (Note 1) An information processing system, characterized in that it comprises: Means for receiving text information containing active content via a communication network through a user terminal and storing the text information in a storage device; This is a method for generating prompts for generative artificial intelligence models. The text information is input into the natural language parsing function of the generative artificial intelligence model. The parsing process extracts business domain, activity type, scale information and multiple feature words from the text information and generates structured data. This is a means of retrieving information storage devices that have registered financial assistance programs based on the structured data, calculating evaluation values ​​for multiple candidate financial assistance programs based on industry suitability, activity suitability, and condition matching, and selecting the optimal financial assistance program based on the evaluation values. A means for obtaining a document template based on the format information contained in the application requirements of the selected financial assistance system, and for automatically assigning values ​​to predetermined items in the document template using the structured data and the text information, in order to generate an initial version of the application document data. Means for generating prompt statements for a generative artificial intelligence model for at least a portion of the purpose description, background description, or effect description in the initial version data of the application document, automatically generating or modifying the explanatory text through the text generation processing of the generative artificial intelligence model, and reflecting the explanatory text in the initial version data of the application document; Means for sending the initial version of the application document data and the reasons for recommendation related to the selected financial assistance system to the user terminal through a user interface, and reflecting the corrections received from the user terminal into the initial version of the application document data to generate the final version of the application document data. Means for converting the final version of the application document data into a predetermined electronic document format or predetermined data structure and sending it to an external review device to complete the application processing.

[0523] (Note 2) The information processing system according to Appendix 1 is characterized in that, The means for generating the final version of application document data is configured to: detect the presence, numerical format, and date format of required fields in the application document data containing the revised content received from the user terminal; when incompleteness is detected, generate indication information corresponding to the incomplete content and send it to the user terminal; and update the final version of application document data based on the application document data re-entered by the user terminal according to the indication information.

[0524] (Note 3) The information processing system according to Appendix 1 is characterized in that, The system further includes: generating prompt statements for a generative artificial intelligence model to generate explanatory text explaining the reasons for the suitability of the selected financial assistance system, sending the prompt statements to the generative artificial intelligence model, and displaying the explanatory text obtained from the generative artificial intelligence model as the recommendation reasons on the user terminal through the user interface.

[0525] Example 2 (Note 1) An information processing system, characterized in that it comprises: A device for receiving project-related information from a user terminal and registering the information into a storage device; An apparatus for constructing prompt statements for a generative artificial intelligence model based on the registered information, sending the information as input data to the generative artificial intelligence model, and obtaining parsing results including domain information and multiple important words; An apparatus for retrieving multiple candidates for financial assistance programs from the storage device based on the domain information and key words in the parsing results, calculating a suitability index for each candidate financial assistance program, and selecting a financial assistance program based on the suitability index. An apparatus for using the application conditions of the selected financial assistance system and the analysis results as input data to construct prompt statements for each component of the application document for a generative artificial intelligence model, and to obtain application document text data containing multiple chapters through the generative artificial intelligence model. An apparatus for automatically embedding the acquired text data into a template of a document generation description format to generate an electronic document of a predetermined format; An apparatus for sending structured text data corresponding to the electronic document to a user terminal via a user interface, and for receiving edits from the user to update the structured text data and the electronic document; A device for storing the updated electronic document as the final version and outputting application processing completion information.

[0526] (Note 2) The information processing system according to Appendix 1 is characterized in that, The apparatus for constructing prompts for a generative artificial intelligence model is configured to generate at least the following distinct prompts: a prompt for extracting domain information, key words, and summary text of funding assistance schemes from the project description; a prompt for calculating suitability based on the project description and multiple funding assistance scheme candidates; and a prompt for generating the text of each chapter of the application document, and to sequentially invoke the generative artificial intelligence model by generating the prompts in stages.

[0527] (Note 3) The information processing system according to Appendix 1 is characterized in that, The document generation description format is a markup language format, the electronic document is a fixed-format file, and the user interface allows browsing the fixed-format file and editing the corresponding structured text data to be performed in parallel.

[0528] Application Example 2 (Note 1) An information processing system, characterized in that it comprises: A means for receiving information related to events and information related to emotional states from a user, and storing the information in a storage device; A device for performing natural language processing on the information related to the matter, and extracting structured information including the type, purpose, scale, budget, period, and target area of ​​the matter, as well as multiple key information from the information; The device is used to retrieve a system information storage device that stores multiple support systems based on the structured information and the multiple key information, extract support system candidates with high adaptability to the matter, calculate an adaptability score for each support system candidate, and select the optimal support system. An apparatus for automatically generating prompt statements that are input into a generative artificial intelligence model based on the structured information and the institutional requirements information of the selected support system, and using the prompt statements to cause the generative artificial intelligence model to perform application document generation processing. An apparatus for adding conditions regarding the tone of the document, the level of detail of the explanation, and the emphasis of items to the prompt statement based on information related to the user's emotional state and the results of emotional analysis processing, thereby generating additional prompt statements for optimizing the application document according to the emotional state, and inputting the additional prompt statements into the generative artificial intelligence model so that the generative artificial intelligence model can automatically adjust the content or expression of the application document; A device for recording the application documents output by the generative artificial intelligence model and converting the application documents into an electronic document format for external submission; And means for transmitting the application document in the electronic document format to an external review device or a support system providing device via a communication device.

[0529] (Note 2) The information processing system according to Appendix 1 is characterized in that, The apparatus for processing information related to a user's emotional state is configured to acquire at least one of the user's input emotion type, voice data, image data, or text data as information related to the emotional state, and during the emotion analysis process, apply an emotion classification model or emotion analysis service to the information to calculate the emotion type and emotion intensity, and control at least one of the following in the prompt statement: the amount of explanation, the reassuring expression, or the concise expression, based on the emotion type and the emotion intensity.

[0530] (Note 3) The information processing system according to Appendix 1 is characterized in that, The system is configured to send the application document generated by the generative artificial intelligence model to a user terminal with a display device, so that the user terminal displays the application document in an editable form, receives corrections from the user, generates a second application document containing the corrections based on the corrections, and sends the second application document to the external review device or support system providing device.

Claims

1. An information processing system, characterized in that, include: processor; The processor is configured as follows: Receive information from the user and save the information to the database; The received information is analyzed using a generative artificial intelligence model, and important keywords are extracted from the information. Based on the extracted keywords, prompts are generated to select the most suitable financial assistance system; It also automatically generates application documents for applying for the selected financial assistance program using prompts generated based on the requirements of the selected financial assistance program.

2. The information processing system according to claim 1, characterized in that, The processor is further configured to analyze the user's emotional state and generate prompts based on the emotional state for adjusting the application documents.

3. The information processing system according to claim 1, characterized in that, The processor is further configured to display the generated application documents through a user interface, and to enable the user to confirm and revise the application documents.

Images